Spacetime Diagrams

Up to this point, special relativity has been a list of effects. Moving clocks run slow. Moving rods are short. Simultaneity is relative. Each is derived from the postulates, each is correct, and each has to be remembered separately — and if you have ever felt that relativity is a bag of tricks with a factor of $\gamma$ stapled to everything, that feeling is honest. It is what the theory looks like when you meet it as a list. In September 1908 Hermann Minkowski stood up in Cologne and proposed that the list was the wrong way to see it. Space and time, he said, are not two things that happen to get mixed up when you go fast. They are one four-dimensional thing — and the effects on the list are not separate phenomena at all. They are what a single geometric object looks like when it is sliced at different angles. From that moment special relativity stops being a collection of results about clocks and rods and becomes something else: the geometry of a four-dimensional manifold with an indefinite metric. This article teaches you to draw the picture. It is a primer, and its job is vocabulary and fluency rather than cleverness — everything here is a tool that the rest of the Minkowski chapter will use without further comment. You will learn why time goes up the page, why we plot $ct$ rather than $t$ , why light travels at exactly $45°$ , and what it means that a moving observer's version of the same diagram looks different in every respect but one. That one exception is the point of the whole article. Almost everything you can draw is frame-dependent: the axes tilt, the grid skews, lengths and durations change. The light cone does not. Every observer, no matter how fast, draws the very same cone through any given event — and that single invariant object turns the diagram from a picture into a causal map.

Key takeaways

A spacetime diagram plots events as points. Time runs up the page; one space direction runs across. A worldline is the curve traced by an object — a history, not a trajectory.
Plot $ct$ (not $t$ ) on the vertical axis. Then both axes are in metres, light travels at exactly $45°$ , and the geometry becomes visible instead of buried in units.
The geometry is not Euclidean. The invariant is $s^2 = (c\Delta t)^2 - \Delta x^2$ — with a minus sign — so "distance" in spacetime behaves in ways your Euclidean eye will actively misread. Trust the algebra over the picture.
OWNED: the light cone is the one feature every observer draws identically. Boosting tilts the axes, skews the grid, and changes every coordinate on the page — but the two null rays through an event map to themselves. That is exactly the statement that $s^2 = 0$ is invariant.
OWNED: the cone is a field, not a feature of the origin. There is a cone at every event. Intersect the future cone of $A$ with the past cone of $B$ and you get the causal diamond — everything that $A$ can influence and that can influence $B$ . For $A = (0,0)$ and $B = (ct = 10, x = 0)$ m, it is a square with corners at $(5, \pm5)$ and area $50\ \mathrm{m}^2 = s^2/2$ , the same in every frame.
The diamond is exactly the set of turning points available to a worldline from $A$ to $B$ . Turn at $x = 3$ and you age $8\ \mathrm{m}/c$ instead of $10$ ; turn at the corner $x = 5$ and you age nothing at all.
The straight worldline between two events has the greatest proper time — the opposite of Euclid. We use that theorem here and draw it; it is proved in Proper Time.
Key idea: a spacetime diagram is a map of causality, not a picture of space. Read it by looking at the cones, because the cones are the only thing on the page that every observer agrees about.

The core insight

Draw time up and space across, and measure time in metres by plotting $ct$ . Then light moves on lines at exactly $45°$ , and those lines — the light cone — are the only marks on the page that every inertial observer draws in exactly the same place. Everything else (the axes, the grid, which events are simultaneous, how long a metre is) tilts when you change frames. So read the diagram by its cones: they carve spacetime at every event into a future you can reach, a past that can reach you, and an elsewhere that can do neither.

Where this lives today

The spacetime diagram is not a teaching aid that working physicists outgrow; it is the working notation of several fields at once. Every Feynman diagram is a spacetime diagram with the axes left implicit — the vertices are events, the lines are worldlines, and the reason antiparticles can be drawn as particles running backwards in time is a statement about the cone structure. In general relativity the whole subject is the study of what happens when the cones stop being parallel: tilt them and you get gravity, tilt them far enough and the future cones of every event point inward and you have crossed an event horizon. The Penrose diagram — the standard tool for reasoning about black holes — is a spacetime diagram deformed so that light stays at $45°$ even at infinity, precisely because the cone is the one structure worth preserving. Causal diamonds in particular have become central rather than decorative. The entropy bounds of holography are stated in terms of them; in quantum field theory the algebra of observables is attached to diamond-shaped regions, because the diamond is exactly the region an experimenter with a bounded lifetime can prepare and then measure. And the causal-set programme takes the radical step of proposing that the causal order — the cone structure, and nothing else — is spacetime, with the geometry emerging from it. The picture Minkowski drew to tidy up Einstein's kinematics has turned out to be the most durable object in the subject.

[Continue with quiz blocks]

The physical picture

Start with the primitive object. An event is a point in spacetime: a place and a time, taken together. A firecracker going off at the end of the platform at noon is an event. So is a particular tick of a particular clock. An event is not a thing; it is a happening, with no duration and no extent — the spacetime equivalent of a point in geometry. A worldline is the curve an object traces out. Here is where most people's intuition needs adjusting: a worldline is not a path through space. It is a history. Even an object sitting perfectly still has a worldline, because it is still moving through time — its worldline is a vertical line. A stationary rock, over the course of a year, sweeps out a worldline a light-year long. Nothing in spacetime is ever at rest, because everything ages. Now the drawing conventions, and each one earns its place.

Time goes up. This is pure convention, but it is universal, and it means the future is at the top of the page. Fine.

Plot $ct$ , not $t$ . This one is not convention — it is the single most useful decision in the whole subject, and it is worth being clear about why. If you plot seconds up and metres across, then a light ray has a slope of $3 \times 10^8$ and is indistinguishable from the vertical axis; the interesting geometry is squashed into an invisible sliver. But $c$ is a conversion factor between the units we historically used for time and the units we used for space, exactly as $2.54$ converts inches to centimetres. Multiply time by $c$ and you have a length. Now both axes are in metres, and one metre of $ct$ is the time light takes to travel one metre — about $3.34$ nanoseconds. Light now travels on lines of slope $\pm 1$ : exactly $45°$ . The geometry becomes visible. Once light is at $45°$ , everything else follows from one rule: no worldline can ever be more horizontal than $45°$ , because that would mean travelling faster than light. Massive objects always have worldlines steeper than $45°$ . A slow object is nearly vertical. A fast one leans over. The tilt of a worldline is its velocity, and the $45°$ line is the wall it can never reach.

Figure

The anatomy of a spacetime diagram. Time (as ct, in metres) runs UP; one space direction runs across. An EVENT is a point; a WORLDLINE is a history. Because we plot ct, light travels at exactly 45°, and no worldline may ever lean past 45° — the shaded wedges are unreachable. The slow object's worldline is nearly vertical; the fast one (β = 0.6) leans, but not to 45°

The last thing to un-learn is the most stubborn. The page is not Euclidean. Your eye measures distance on paper with Pythagoras, $\sqrt{\Delta x^2 + \Delta y^2}$ , and it will do this to a spacetime diagram whether you ask it to or not. But the quantity that actually matters here has a minus sign in it. Two events that look far apart on the page can have zero separation in spacetime. A worldline that looks longer can correspond to less elapsed time. This is not a flaw in the drawing; it is the geometry being genuinely different from the one your eye was trained on. Draw the picture, but do the arithmetic.

The core equations

There is really only one, and everything in the chapter is a consequence of it.

s^2 = (c\,\Delta t)^2 - \Delta x^2 - \Delta y^2 - \Delta z^2 \qquad (1)

where: $s^2$ is the invariant interval between two events; $\Delta t, \Delta x, \dots$ are their coordinate differences. Signature $(+,-,-,-)$ , timelike-positive: $s^2 > 0$ is timelike, $s^2 = 0$ is lightlike (null), $s^2 < 0$ is spacelike. Physical reading: this is the spacetime replacement for Pythagoras, and the minus signs are the entire difference between this geometry and Euclid's. Every observer computes a different $\Delta t$ and a different $\Delta x$ for the same pair of events — and every observer gets the same $s^2$ . It is derived and proved invariant in The Invariant Spacetime Interval; here we simply use it. The three signs of $s^2$ are the three ways two events can be related, and the whole causal structure of the diagram is nothing but a reading of that sign.

s^2 > 0 \;\;\text{TIMELIKE} \qquad s^2 = 0 \;\;\text{LIGHTLIKE} \qquad s^2 < 0 \;\;\text{SPACELIKE} \qquad (2)

where: the classification is frame-independent, because $s^2$ is. Physical reading: timelike means a massive object can get from one event to the other — they can be cause and effect, and some observer sees them at the same place. Lightlike means only a light ray can make the trip. Spacelike means nothing can: no signal connects them, neither can cause the other, and some observer sees them as simultaneous. The classification is owned by Timelike, Spacelike and Lightlike Separations; what this article adds is what it looks like on the page. Setting $s^2 = 0$ gives the boundary between "can influence" and "cannot" — and in the $(x, ct)$ plane that boundary is a pair of straight lines.

(c\,\Delta t)^2 - \Delta x^2 = 0 \quad\Longleftrightarrow\quad c\,\Delta t = \pm\,\Delta x \qquad (3)

where: the two solutions are the two null rays. Physical reading: these are the $45°$ lines. In a full three-space they sweep out a cone — hence the name, and we keep the word "cone" even in the one-space-dimension drawings where it is really just an X. Because $s^2$ is invariant, the set of events with $s^2 = 0$ from a given event is the same set for every observer. This is the article's first owned result, and it is the reason the diagram is worth drawing at all. Finally, the aging of a clock along a worldline — quoted here, not derived.

\tau = \int \sqrt{1 - \beta^2(t)}\;dt \qquad\text{and}\qquad \tau_{\text{straight}} > \tau_{\text{any other path}} \qquad (4)

where: $\tau$ is the proper time — what a clock carried along the worldline actually reads — and $\beta(t) = v(t)/c$ . Physical reading: (4) is proved in Proper Time, which owns both the integral and the theorem, and we take both as given. The theorem is the exact opposite of the Euclidean fact you are used to: in a plane, the straight line is the shortest path; in spacetime, the straight worldline is the one of greatest proper time. The straight line is the way to age the most. Keep this in mind whenever the page tempts you to trust your eye — on the page, the bent path looks longer, and it is precisely the one that ages less.

The light cone, and why it is the only thing everyone agrees on

Derivation AOWNED — the light cone is the one feature every observer draws identically

When you change to a frame moving at $\beta$ , almost everything on the diagram changes. The time axis tilts. The space axis tilts the other way. The grid of "simultaneous" lines skews. Every event's coordinates take new values. It is fair to ask whether anything survives — and if nothing did, the diagram would be a picture of one observer's prejudices rather than a map of reality. Something does survive, and it is exactly one thing. This derivation shows that the two null rays through an event are mapped to themselves by every boost, and that this is not a coincidence but a restatement of the invariance of the interval.

Step 1. Write down what a boost does to the coordinates. In the $(ct, x)$ plane, $c t' = \gamma(c t - \beta x)$ and $x' = \gamma(x - \beta\, ct)$ , with $\gamma = 1/\sqrt{1-\beta^2}$ . Why: this is the Lorentz transformation, derived in The Lorentz Transformation Equations. We only need to apply it.

Step 2. Confirm the interval is unchanged. Substituting and expanding, $(ct')^2 - (x')^2 = \gamma^2\left[(ct - \beta x)^2 - (x - \beta ct)^2\right] = \gamma^2 (1-\beta^2)\left[(ct)^2 - x^2\right] = (ct)^2 - x^2$ , since $\gamma^2(1-\beta^2) = 1$ exactly. Why: this single line is the engine of everything that follows. The cross terms cancel and the leftover factor is precisely the one $\gamma$ was defined to make equal to one.

Step 3. Now ask what happens to the set $s^2 = 0$ . Because $s'^2 = s^2$ for every pair of events, an event on the cone ( $s^2 = 0$ ) has $s'^2 = 0$ too. So the cone maps into the cone — every observer draws the same set of events. Why: this is the whole result already, and note how cheaply it came. We did not need to know anything about light beyond the fact that $s^2 = 0$ is what "connected by a light ray" means.

Step 4. Sharpen it: the rays are not merely mapped into the cone as a set, they are mapped along themselves. Apply the boost to the null direction $(ct, x) = (1, 1)$ : it becomes $\gamma(1 - \beta)\,(1, 1)$ , the very same direction, rescaled by $\gamma(1-\beta) = \sqrt{(1-\beta)/(1+\beta)}$ . Likewise $(1, -1) \mapsto \gamma(1+\beta)(1,-1)$ . The null rays are eigenvectors of every boost. At $\beta = 0.6$ the two factors are $0.5$ and $2$ exactly. Why: this is the sharpest possible form of the statement. It is not that the cone happens to land on top of itself; each ray is individually fixed as a line, merely stretched along its own length. (That the stretch factor $\sqrt{(1+\beta)/(1-\beta)}$ is the relativistic Doppler factor is developed in Lorentz Boosts as Hyperbolic Rotations.)

Step 5. Draw the conclusion the primer needs. On a spacetime diagram, the axes are your axes, the grid is your grid, and the simultaneity lines are your convention. Change observer and all of them move. The $45°$ cone is the one structure drawn identically by everyone. It is therefore the only safe thing to reason with — and, happily, it is the thing that encodes causality. Why: this is why we teach the diagram at all. A picture whose every feature was observer-dependent would be worthless as physics. The cone is what makes it a map. *These results hold for all $\beta$ : $s'^2 = s^2$ symbolically; both null directions are eigenvectors with eigenvalues $\sqrt{(1\mp\beta)/(1\pm\beta)}$ , giving the numerical values $0.5$ and $2$ at $\beta = 0.6$ ; and any event on the cone remains on the cone after boosting.*

Common trapYour Euclidean eye will lie to you about this picture

Three specific lies, all worth naming. (1) "Longer on the page means more time." It means less. The straight worldline from $A$ to $B$ looks like the shortest line in the figure, and it is the one with the greatest proper time. Bend it and it gets visibly longer and ages visibly less. (2) "Those two events look close together, so their separation is small." If they are near the cone, their separation is near zero, however far apart they look. Two events on the same null ray have $s^2 = 0$ exactly, even if the page shows a metre between them. (3) "The axes are perpendicular, so the frame is the right one." Perpendicular-on-the-page is a property of the ink, not of spacetime; a moving observer draws axes at an angle and is exactly as correct as you are. The cure for all three is the same: the picture is for intuition, the interval $s^2$ is for truth. When they disagree, the interval wins.

The cone at every event: causal diamonds

Go deeper · Derivation B · 4 minOWNED — the cone is a field, and it turns the diagram into a causal map

Derivation BOWNED — the cone is a field, and it turns the diagram into a causal map

So far we have drawn one cone, through one event, and it is tempting to think of it as a feature of the origin — a special place on the page. It is not. Nothing distinguishes the origin: spacetime is homogeneous, and the construction of Step 1 above works at any event. There is a light cone at every event, and once you see the diagram as carrying a whole field of cones, you can start intersecting them — and that is where the diagram stops being a picture and starts answering questions.

Step 1. Put a cone everywhere. Choose any event $E$ ; the set of events with $s^2 = 0$ from $E$ is a pair of $45°$ rays through $E$ . Nothing in the derivation referred to the origin. Why: this is the geometric content of homogeneity — no event is special. The diagram is not a picture with a cone in it; it is a picture that is cones, all the way across.

Step 2. Define the two halves. The future cone of $E$ , written $J^+(E)$ , is everything $E$ can influence: all events with $s^2 \geq 0$ and $\Delta t > 0$ . The past cone $J^-(E)$ is everything that can influence $E$ . Everything else is elsewhere — spacelike separated, causally disconnected in both directions. Why: the sign of $s^2$ decides whether a signal can connect the events, and the sign of $\Delta t$ decides which way. The second is frame-independent too, but only for timelike and null pairs — that is the content of Timelike, Spacelike and Lightlike Separations.

Step 3. Intersect two cones. Take two timelike-separated events, $A$ (earlier) and $B$ (later). The set $J^+(A) \cap J^-(B)$ — everything $A$ can influence and which can in turn influence $B$ — is called the causal diamond of $A$ and $B$ . In the $(x, ct)$ plane it is literally a diamond: bounded above by $B$ 's past cone, below by $A$ 's future cone. Why: this is the set of everything that can "get a message from $A$ to $B$ through it". It is the complete answer to "what could have mediated between these two events?" — and it is a bounded region, which is the surprise. Causality between two events is a local affair.

Step 4. Compute it. Take $A = (ct, x) = (0, 0)$ and $B = (10, 0)$ metres. $A$ 's future cone is $ct \geq |x|$ ; $B$ 's past cone is $10 - ct \geq |x|$ . The two conditions meet at $ct = 5$ , $x = \pm 5$ . The diamond has vertices $(0,0)$ , $(5, 5)$ , $(10, 0)$ , $(5, -5)$ — a square standing on its corner, with diagonals of $10\ \mathrm{m}$ each. Its area is $\tfrac12 d_1 d_2 = \tfrac12 \times 10 \times 10 = \mathbf{50\ m^2}$ . And the interval between $A$ and $B$ is $s^2 = 10^2 - 0^2 = 100\ \mathrm{m}^2$ , so the area is exactly $s^2/2$ . Why: the diamond's size is set by the interval and nothing else — which already suggests the area is an invariant, since $s^2$ is.

Step 5. Prove the diamond is frame-independent, in two senses. First, as a set: it is built entirely out of light cones, and Step 4 of the previous derivation showed every observer draws the same cones — so every observer picks out the same events as "the diamond". Second, its area is invariant: a boost has $\det \Lambda = 1$ , so it preserves areas in the $(x, ct)$ plane (this is proved in Minkowski Diagrams and Skewed Coordinate Axes). Boosting at $\beta = 0.6$ carries the four vertices to $(0,0)$ , $(2.5, 2.5)$ , $(12.5, -7.5)$ , $(10, -10)$ — a visibly different, sheared-looking quadrilateral — whose area is still exactly $\mathbf{50\ m^2}$ , and whose $s^2$ between the new $A$ and $B$ is still $12.5^2 - 7.5^2 = \mathbf{100\ m^2}$ . Why: the shape on the page changes completely and the physics does not change at all. This is the single best illustration in the chapter of what "frame-dependent" versus "invariant" actually means.

Step 6. Now the payoff — read the diamond as a set of choices. Suppose you travel from $A$ to $B$ and are allowed one course change. Where may you turn? Exactly at the events of the diamond: any turning point outside it would require a leg faster than light. The causal diamond is precisely the set of turning points available to a worldline from $A$ to $B$ . And the proper time you accumulate depends on where you turn: turning at $(5, x)$ gives $\tau = 2\sqrt{5^2 - x^2}$ . At $x = 0$ — no turn, the straight worldline — that is $\tau = 10\ \mathrm{m}/c$ , the maximum. At $x = 3$ : $\tau = 2\sqrt{16} = 8\ \mathrm{m}/c$ . At $x = 4$ : $\tau = 6\ \mathrm{m}/c$ . At the diamond's corner, $x = 5$ : $\tau = 0$ . Why: this ties the whole article together. The theorem that the straight path maximises proper time (proved in Proper Time) is not an abstract fact here — you can watch $\tau$ fall to zero as the turning point slides out to the corner of the diamond, where the path is two null rays and the traveller does not age at all. The corner of the diamond is a light beam's itinerary, and light does not age. *These results can be confirmed directly: the four vertices follow from the two cone inequalities; the area $= 50\ \mathrm{m}^2$ by the shoelace formula, equal to $s^2/2$ ; the boosted vertices at $\beta = 0.6$ yield area still $50\ \mathrm{m}^2$ (shoelace); $s^2 = 100\ \mathrm{m}^2$ before and after the boost; and $\tau(x) = 2\sqrt{25 - x^2}$ at $x = 0, 3, 4, 5$ gives $10, 8, 6, 0$ m $/c$ .*

Figure

PUNCHLINE — the cone is a FIELD, and two cones make a causal diamond. A = (0,0), B = (ct = 10, x = 0) m, so s² = 100 m². The future cone of A and the past cone of B intersect in a diamond with corners (5, ±5) and area 50 m² = s²/2. That diamond is exactly the set of TURNING POINTS available to a worldline from A to B — and where you turn decides how much you age: straight ⇒ τ = 10 m/c, turn at x = 3 ⇒ τ = 8 m/c, turn at the CORNER x = 5 ⇒ τ = 0. Light does not age

Figure

The same diamond, drawn by an observer moving at β = 0.6 — and this is what "invariant" means. Every vertex has moved: A = (0,0), the corner (5,5) → (2.5, 2.5), B = (10,0) → (12.5, −7.5), the corner (5,−5) → (10, −10). The shape on the page is unrecognisable. But it is the SAME SET of events, the light cone is still at 45°, s²(A,B) = 12.5² − 7.5² = 100 m² still, and the area is STILL exactly 50 m². The ink changes; the physics does not

Limiting cases

Case	What the diagram does	Meaning
$\beta \to 0$	axes upright, worldline vertical	you are at rest; this is your frame, and it is not special
$\beta \to 1$	worldline leans toward $45°$	the cone is a wall, approached but never reached by anything with mass
$c \to \infty$	the cone opens out flat	the Galilean limit: "elsewhere" vanishes, every event is in your future or past, and simultaneity becomes absolute
$s^2 \to 0$	events slide onto the cone	separation vanishes although the page still shows a gap — the classic Euclidean lie
diamond, $B \to A$	the diamond shrinks to a point	two events with nothing between them; area $s^2/2 \to 0$
corner of diamond	$\tau \to 0$	the null path: light traverses it and does not age

Worked examples

1 · Read the causal structure straight off a diagram

Event $A$ is at $(ct, x) = (0, 0)$ . Classify each of $P = (10, 4)$ , $Q = (4, 10)$ , $R = (7, 7)$ metres, and say what each means physically.

Step 1. Compute the interval for each — this is the only calculation there is. $s^2 = (c\Delta t)^2 - \Delta x^2$ .

$P$ : $s^2 = 100 - 16 = \mathbf{+84\ m^2} > 0$ — timelike.
$Q$ : $s^2 = 16 - 100 = \mathbf{-84\ m^2} < 0$ — spacelike.
$R$ : $s^2 = 49 - 49 = \mathbf{0}$ — lightlike.

Step 2. Translate. $P$ is inside $A$ 's future cone: a rocket can get from $A$ to $P$ (it needs $\beta = 4/10 = 0.4$ , comfortably under $1$ ), so $A$ could have caused $P$ . $Q$ is elsewhere: reaching it from $A$ would need $\beta = 10/4 = 2.5$ , so no signal connects them and neither can cause the other — and some observer sees them as simultaneous. $R$ lies exactly on the cone: only a light ray makes the trip, arriving with nothing to spare.

Step 3. Check the boundary number. The dividing case is always $|\Delta x| = c\Delta t$ , i.e. $\beta = 1$ . Everything with $|\Delta x| < c\Delta t$ is reachable; everything with $|\Delta x| > c\Delta t$ is not. Notice that $P$ and $Q$ have $s^2$ of equal magnitude and opposite sign — they are mirror images across the cone, and they could hardly be more different physically. The sign of $s^2$ is the physics; its magnitude is the detail.

2 · The diamond, and where you are allowed to turn

You leave event $A = (0,0)$ and must arrive at $B = (ct = 10, x = 0)$ m. You may change course exactly once. Where may you turn, and how much do you age?

Step 1. Find the permitted region. You must stay inside $A$ 's future cone (you cannot outrun light on the way out) and inside $B$ 's past cone (you must still be able to get back). That intersection is the causal diamond: corners $(0,0)$ , $(5,5)$ , $(10,0)$ , $(5,-5)$ , area $\mathbf{50\ m^2} = s^2/2$ .

Step 2. Turn at the mid-time, $ct = 5$ , at position $x$ . Each leg is a straight timelike segment with $\Delta(ct) = 5$ and $\Delta x = \pm x$ , so each contributes $\sqrt{25 - x^2}$ and the total is $\tau(x) = 2\sqrt{25 - x^2}$ metres of $c\tau$ .

Step 3. Tabulate, and read the moral.

turn at $x$	$\tau$ (m of $c\tau$ )	in seconds	comment
$0$	$\mathbf{10.000}$	$0.03336\ \mu\mathrm{s}$	no turn at all — the straight worldline. Maximum
$2$	$9.165$	$0.03057\ \mu\mathrm{s}$	a gentle excursion already costs you
$3$	$\mathbf{8.000}$	$0.02668\ \mu\mathrm{s}$	the path looks longer on the page and ages less
$4$	$6.000$	$0.02001\ \mu\mathrm{s}$
$5$	$\mathbf{0.000}$	$0$	the diamond's corner: two null rays. You do not age at all

Step 4. Interpret. The straight path wins, and it wins by the theorem proved in Proper Time — the inertial worldline has the greatest proper time of any path between two events. What the diamond adds is where the competitors live: every rival path must turn inside this region, and as the turning point slides toward the corner your elapsed time falls smoothly to zero. The traveller who turns at the corner is a light beam. This is the twin paradox, already fully visible, with no paradox in sight — the twin who turns around simply took a path of shorter proper time, and you can read that off the picture.

Physics of the portrait

Origins · 2 minHermann Minkowski and how this was discovered

Hermann Minkowski

1864–1909 · "Raum und Zeit" (1908) · "Eigenzeit" · the four-dimensional formalism

Minkowski was a pure mathematician — a specialist in number theory and the geometry of numbers — who had taught the young Albert Einstein mathematics at the ETH in Zurich. (A much-repeated story has him calling his former student a "lazy dog"; it is a reported anecdote, not a documented one, and should be carried lightly.) In 1907 he read a paper, "Die Grundgleichungen für die elektromagnetischen Vorgänge in bewegten Körpern", which recast electrodynamics in four-dimensional form and introduced the term "Eigenzeit" — proper time. And on 21 September 1908, at the 80th Assembly of German Natural Scientists and Physicians in Cologne, he delivered the lecture that named the subject. "Von Stund an sollen Raum für sich und Zeit für sich völlig zu Schatten herabsinken" — henceforth space by itself, and time by itself, are doomed to fade away into mere shadows, and only a kind of union of the two shall preserve an independent reality. The claim was not that Einstein's results needed fixing; they were correct. It was that they were being stated in the wrong language — as a list of things that go wrong at high speed, rather than as the geometry of a single four-dimensional object. He was right, and the reformulation turned out to be far more than cosmetic: it is what made general relativity thinkable, since you cannot curve a spacetime you have not yet noticed you have. Minkowski did not live to see any of it. He died on 12 January 1909 of a ruptured appendix, four months after the lecture, aged 44.

History

1905

— Einstein derives time dilation, length contraction and the relativity of simultaneity from the two postulates, and states the theory entirely without geometry.

1907

— Minkowski reads "Die Grundgleichungen…" (published 1908), giving the four-dimensional formalism and coining "Eigenzeit", proper time.

1908

— On 21 September, at the 80th Assembly of German Natural Scientists and Physicians in Cologne, Minkowski delivers "Raum und Zeit"; it appears in Physikalische Zeitschrift 10, 104–111, in 1909.

1909

— On 12 January, Minkowski dies of a ruptured appendix, four months after the lecture.

1915

— Einstein completes general relativity, in which Minkowski's flat spacetime becomes the local approximation to a curved manifold.

1966

— Taylor and Wheeler publish Spacetime Physics (W. H. Freeman), which makes the diagram the standard first tool of the subject rather than an afterthought.

Going further

Texts. Taylor & Wheeler, Spacetime Physics, 2nd ed., Ch. 1–3 — still the best introduction to the diagram · Minkowski, "Raum und Zeit", Physikalische Zeitschrift 10 (1909) 104–111; translated in The Principle of Relativity (Dover) · Rindler, Introduction to Special Relativity, 2nd ed., §8 · Misner, Thorne & Wheeler, Gravitation, Ch. 1, for the cone-as-field view. The diagram, the $45°$ convention, and the three-way causal classification are standard and are developed in The Invariant Spacetime Interval and Timelike, Spacelike and Lightlike Separations. The two results this article develops in full are, first, the cone-invariance result — that the light cone is the unique feature every observer draws identically, sharpened to the statement that the two null rays are eigenvectors of every boost with eigenvalues $\sqrt{(1\mp\beta)/(1\pm\beta)}$ (which are $0.5$ and $2$ at $\beta = 0.6$ ) — and, second, the causal diamond: the cone as a field, the diamond as the intersection $J^+(A) \cap J^-(B)$ , its area $= s^2/2 = 50\ \mathrm{m}^2$ invariant under boosts, and the reading of the diamond as exactly the set of turning points available to a worldline from $A$ to $B$ , with $\tau(x) = 2\sqrt{25 - x^2}$ falling from $10\ \mathrm{m}/c$ at the centre to $0$ at the corner. The theorem that the straight worldline maximises proper time is used, drawn, and attributed — it is proved in Proper Time, not here. Signature: $(+,-,-,-)$ , timelike-positive — note that many texts use the opposite convention $(-,+,+,+)$ , in which every $s^2$ flips sign; the physics is identical, the bookkeeping is not.

Builds on. Events and Worldlines · The Invariant Spacetime Interval. Leads to. The Light Cone · Invariant Hyperbolae · Causal Structure · The Hyperbolic Geometry of Boosts.

Why is it called a light "cone" when the diagram just shows an X?

Because the diagram is showing you a two-dimensional slice of a four-dimensional object. With one space dimension, the set of events reachable by light from here-and-now is two rays — an X. Restore a second space dimension and those rays sweep around into an actual cone, opening upward into the future and downward into the past. With all three space dimensions it is a three-sphere of light expanding at $c$ , which nobody can draw, and "cone" is what we keep calling it. The word survives because the structure survives: a boundary between the reachable and the unreachable, with $s^2 = 0$ on it.

If a moving observer draws different axes, is my diagram wrong — or is his?

Neither, and the question is the same mistake as asking whether your street map is wrong because someone else oriented theirs with east at the top. Both diagrams contain exactly the same events and exactly the same cones. What differs is only the coordinate grid laid over them, and a grid is a bookkeeping choice, not a physical fact. The two of you will disagree about which events are simultaneous, about how long a metre is, and about how fast a clock ticks — and you will agree, exactly, about every interval $s^2$ , about every light cone, and about every physical coincidence (this clock read that as it passed this marker). The disagreements are about the ink; the agreements are about the world. How the two grids relate is worked out in Minkowski Diagrams and Skewed Coordinate Axes.

The bent worldline is obviously longer. How can it possibly show LESS elapsed time?

Because "longer" is your Euclidean eye talking, and the eye is measuring the wrong thing. Length on the page is $\sqrt{\Delta x^2 + \Delta (ct)^2}$ ; elapsed time along a worldline is $\sqrt{\Delta(ct)^2 - \Delta x^2}$ . Those differ by a sign, and the sign is the whole of relativity. Moving sideways in space (adding $\Delta x$ ) makes the ink longer and the clock reading shorter, because $\Delta x$ enters with a minus. That is why the straight worldline — which spends none of its budget on $\Delta x$ — is the one that ages the most, and why the null path around the corner of the diamond ages not at all. If you want a slogan: in spacetime, the shortest way between two events is to take the scenic route.

Is the causal diamond a real region, or just a bookkeeping device?

It is as real as anything in the diagram, in the precise sense that all observers agree on which events belong to it — it is constructed entirely out of light cones, and the cones are invariant. It also answers a real question with a real answer: if $A$ influenced $B$ , then whatever mediated between them happened inside the diamond, and nowhere else. That has teeth. It means causal influence between two events is confined to a bounded region of spacetime, not smeared over the universe, which is why the diamond has become the natural home for local observables in quantum field theory and for entropy bounds in holography.

Does light really "not age"? What does τ = 0 actually mean?

It means that a null path accumulates zero proper time: $s^2 = 0$ along every segment of it, so the integral of $\sqrt{s^2}$ over the path is zero. Be careful about what that licenses you to say. It is not a statement about a photon's private experience — you cannot ride along with a light beam, because there is no inertial frame in which light is at rest, and the limiting procedure that would take you there is singular rather than merely awkward. The safe and complete statement is the one in the diagram: a clock carried along a null path would read the same at the end as at the start, and the corner of the causal diamond is exactly where that happens.

My textbook writes s² = −c²Δt² + Δx² and gets all the signs backwards from yours. Who is right?

Both of us, and this trips up more people than any other single thing in relativity. There are two conventions in use: $(+,-,-,-)$ — used throughout this article, in which timelike separations have $s^2 > 0$ — and $(-,+,+,+)$ , in which timelike separations have $s^2 < 0$ . They differ by an overall sign and nothing else, and every physical statement comes out the same, provided you never mix them within one calculation. The one that bites is exactly the one you noticed: whether "timelike" means positive or negative. This article and the whole of this chapter use $(+,-,-,-)$ , timelike-positive. If you are reading a source that does not, negate every $s^2$ before comparing, and check what it means by "timelike" before trusting a single sign.

Spacetime Diagrams

Where this lives today

The physical picture

The core equations

The light cone, and why it is the only thing everyone agrees on

The cone at every event: causal diamonds

Limiting cases

Worked examples

1 · Read the causal structure straight off a diagram

2 · The diamond, and where you are allowed to turn

Physics of the portrait

Going further

Before this

Where this leads