Chapter 6

The Interval: Hero of Spacetime

The one thing every observer agrees on, with a minus sign.

About 60 minutes · Builds on Chapter 5

What this chapter does

Lin and Rae disagree about when and where things happen. This chapter finds the one number they always agree on. It is the hero of the whole course: every later chapter is this number seen from a new side.

  1. Find the number (about 15 minutes). Two notebooks, one pair of firecrackers, and two linked labs: turning a map next to boosting a rocket.
  2. See its picture (about 20 minutes). On a map, equal distances make a circle. In spacetime, one minus sign turns that circle into a hyperbola. The same sign sorts every event into future, past or elsewhere.
  3. Use it (about 25 minutes). Practice, three small functions, and a checkpoint.

By the end you can say what the minus sign does, and why no rocket can turn an effect into a cause.

Warm-up

Three quick questions from earlier chapters. Each one is an ingredient of today’s hero.

From Chapter 1. On the Daytimers’ map, the well is 300 m east and 400 m north of the town square. How far is the well from the square? Need a refresher?

Hint

Square each offset, add, then take the square root.

Solution

\(\sqrt{300^2 + 400^2} = \sqrt{250\,000} = 500\un{m}\). Both guilds got 500, whichever way their north pointed.

From Chapter 4. A flash leaves \(x = 0\) at \(t = 0\) and heads right. Which event does the light pass through?

  • t = 5, x = 3
  • t = 5, x = 5
  • t = 3, x = 5
Solution

Light travels 1 m per meter of time, so it reaches \(x = 5\) at \(t = 5\): the event \((5, 5)\), on the 45° line.

From Chapter 5. Rae’s light clock rides past Lin at \(v = 0.6\). Rae counts 4 m of time per tick. How long is each of those ticks, as Lin’s lattice of clocks measures it?

Hint

Half a tick is a 3-4-5 triangle: legs 1.5 and 2, slant 2.5.

Solution

\(\g = 1.25\) at \(v = 0.6\), so Lin measures \(1.25 \times 4 = 5\). Keep that 5 in mind: the same 5 is waiting in the puzzle below.

The puzzle

Remember the surveyors? Two guilds mapped one town with two different norths. They argued about every east and every north, and agreed perfectly on every distance.

Now the surveyors bring clocks. Two firecrackers go off, at events O and B. Lin, in the lab, writes B down as \((t, x) = (5, 3)\): 5 m of time after O and 3 m to the right. Rae’s rocket moves at 3/5 of light speed, and it passes right through both firecrackers. So for Rae, both bangs happen at her own place, \(x' = 0\). She writes the same pair as \((4, 0)\).

Lin’s notebook Lin, in the lab, writes: O at t 0, x 0; B at t 5, x 3. Lin’s notebook the lab O: t = 0, x = 0 B: t = 5, x = 3 Rae’s notebook Rae, on a rocket moving at three fifths of light speed, writes for the same two firecrackers: O at t′ 0, x′ 0; B at t′ 4, x′ 0. Rae’s notebook rocket, v = 3/5 O: t′ = 0, x′ = 0 B: t′ = 4, x′ = 0
Same two firecrackers, two honest notebooks. Every number disagrees.

Predict

Lin wrote \((5, 3)\) and Rae wrote \((4, 0)\) for the same pair. Which combination of \(t\) and \(x\) do they agree on? Work each option out for both notebooks, on paper or in your head, then lock in your pick. The lab will then test your winner against every other rocket, not only Rae’s.

  • t + x
  • t² + x²
  • t² − x²
  • t − x

Experiment

Two labs, one gesture. The first is the surveyors’ town: turning the Nighttimers’ grid changes every east and every north. The second is a spacetime diagram with O and B. Its speed slider redraws everything as a rocket moving at that speed would measure it. Physicists call that a boost.

The two sliders are tied: move one and the other follows, so a turn and a boost happen side by side. We linked them by hand, so that the special turn (about 53°) and the special boost (\(v = 0.6\)) arrive together. No law of nature connects an angle to a speed. The bar above the labs keeps score. Watch what changes, and what doesn’t.

Map turned 0.0° well (300, 400) east² + north² = 500² circle

The surveyors’ town map from Chapter 1. Turning the Nighttimers’ grid changes the well’s east and north offsets, from (300, 400) to (500, 0) when their east axis points straight at it. The well stays on the teal circle of points 500 m from the square: east² + north² = 500² for both guilds.

Turn the map. Rotate the Nighttimers’ grid: every east and north changes, but the well stays on its teal circle, 500 m from the square.
Describe this lab

A town map with two survey grids: the Daytimers’ grid in solid blue and the Nighttimers’ grid in dashed vermilion, rotated by an angle you set with a slider. A teal circle around the town square passes through the well: every point on it is 500 m from the square. A panel lists the well’s east and north offsets for both guilds and the distance from the square, which is the same for both. This slider is tied to the speed slider of the boost machine: turning the map boosts the rocket.

Try it

Turn the map until the Nighttimers’ east axis points straight at the well. Measure: what angle is that, and what two numbers do the Nighttimers write for the well? Check the bar above the labs: did east² + north² change?

Event B on its hyperbola A spacetime diagram with O at the origin and B at t 5, x 3. A teal hyperbola through B marks every event with t squared minus x squared equal to 16. Rae’s worldline, the dashed line x = 0.6 t, also passes through B. O B t² − x² = 16 t x

B sits at (5, 3) on the teal hyperbola t² − x² = 16. Boosting slides it along that curve: at v = 0.6 it lands at (4, 0), at v = 0.8 at (4.33, −1.67). The panel would list Δt′, Δx′ and τ², which stays 16.

Boost machine. Drag the speed: B slides along its own teal curve, the bold one, and the gold light lines stay put. You can drag B too.
Describe this lab

A spacetime diagram with time up and space to the right, and O’s light cone drawn in gold at 45 degrees. Event O sits at the origin and event B, which you can drag, starts at t 5, x 3. A bold teal hyperbola through B marks every event with the same t² − x² as B; fainter teal hyperbolas mark other values. A speed slider (tied to the map’s rotation) and four speed buttons redraw every event as a rocket at that speed measures it; B slides along its bold curve, and the scale of the plot never changes. If a boost carries B off the plot, a note under it gives B’s coordinates. A panel compares O with B: Δt and Δx in the lab, Δt′ and Δx′ in the rocket, the squared interval τ² and its class (timelike, lightlike or spacelike), and whether O can influence B, and why.

Try it

Press 0.6 under the diagram, or drag the speed. Boosting redraws everything in the rocket’s coordinates: Lin’s axes become the tilted blue lines. Measure: B’s \(t'\) and \(x'\) (Δt′ and Δx′ in the O → B panel). Then press 0.8. Does B ever leave its bold curve? Does \(t'^2 - x'^2\) change?

Try it

Press Reset on the boost machine, then drag B to \((3, 5)\): 3 m of time after O, but 5 m away. Its bold curve now lies on its side. Boost until B’s \(t'\) reads 0. Measure: what speed is that? Keep going past it. Which happens first now, O or B?

Play a little more. Drag B onto a gold 45° line, say \((4, 4)\), and boost. It slides, but it never leaves the line. What does the panel say about \(\tau^2\) there?

Sorting spacetime

Every event around O belongs to one of three families. One number decides which: τ² = \(t^2 - x^2\), worked out from O, the same τ² as in the boost machine’s O → B panel.

  • Positive: timelike. Time wins, and the event is inside O’s light cone.
  • Zero: lightlike. A tie, and the event is on the cone.
  • Negative: spacelike. Space wins, and the event is outside the cone.

Sort eight events into the bins. When all eight are in, a speed slider unlocks.

Eight events around O Event O at the center with its gold light cone. A at t −2, x −2 and C at t 1, x 1 lie on the cone. B at t −3, x 0 is in the past. D at t 3, x 2 and G at t 2, x −1 are in the future. E at t −2, x −4, F at t 1, x −4 and H at t 2, x 4 are outside the cone. O A B C D E F G H

Without the lab: D (3, 2) and G (2, −1) are timelike future; B (−3, 0) is timelike past; A (−2, −2) and C (1, 1) are lightlike; E (−2, −4), F (1, −4) and H (2, 4) are spacelike.

The cone sorter. Drag each event into its bin, tap a card and then a bin, or focus a card and press 1–4.
Describe this lab

Eight lettered events around event O, each shown as a dot on a spacetime diagram and as a card listing its t and x. Four bins: timelike future, timelike past, lightlike, and spacelike or elsewhere. Putting a card in the wrong bin sends it back with a reason. When all eight are sorted, a speed slider unlocks: the diagram is redrawn in a rocket frame, every event slides along its own hyperbola, every card keeps its bin, and spacelike events whose time order with O has flipped are marked.

Try it

Sort all eight. Then boost to \(v = 0.6\), and to \(-0.6\). Measure: on any card, is the rocket’s \(t'^2 - x'^2\) the same as the lab’s \(\tau^2\)? Which events changed their time order with O, and which bin are they all in?

Then play. Hide the light cone and sort by the numbers alone, or press “Shuffle new events” for a fresh set.

What just happened?

Only \(t^2 - x^2\) agrees: Lin gets \(5^2 - 3^2 = 16\), and Rae gets \(4^2 - 0^2 = 16\).

The others don’t. Lin and Rae get 8 and 4 for \(t + x\), 34 and 16 for \(t^2 + x^2\), and 2 and 4 for \(t - x\).

Turning the map changed every east and north, but never east² + north². Boosting changed every \(t\) and \(x\), but never \(t^2 - x^2\). It’s the same game, with one sign flipped.

We call \(t^2 - x^2\) the squared interval and write it \(\inv{\tau^2}\), said “tau squared”. When \(\tau^2\) is positive, its square root \(\tau\) is the interval itself: for O and B, \(\tau^2 = 16\) and \(\tau = 4\). \(\tau^2\) is invariant: every free-float frame gets the same value, and so the same \(\tau\). It is the hero of this course. Time stretching, shrinking rulers, the twins: each is this one number, seen from another side. Even \(E = mc^2\) comes from the same minus sign (Chapter 14 shows how).

The sorter told the same story in another way. If \(\tau^2\) is the same for everyone, then so is its sign. So no boost can ever move a card to another bin. Sagredo’s question is a good one, and the “Why, really?” box below answers it.

One number everybody agrees on

Numbers first

Line up the two stories. The Daytimers put the well at (300, 400). Turn the Nighttimers’ east axis until it points straight at the well, and they put it at (500, 0). Different pairs, same sum of squares: \(300^2 + 400^2 = 500^2 + 0^2\).

Lin puts B at (5, 3). Rae, whose rocket passes through both firecrackers, puts it at (4, 0). Different pairs, same difference of squares: \(5^2 - 3^2 = 4^2 - 0^2\).

Rae’s pair is special, just like the Nighttimers’ was. She is present at both events, so for her the 4 is a plain clock reading: her wristwatch ticks off 4 m of time between O and B. That is the wristwatch time of Chapter 5. Physicists call it proper time, \(\tau\). So the interval \(\tau\) has a meaning you can hold in your hand: it is the time on the watch of anyone who rides straight from one event to the other.

The interval can do real work on its own, with no light clock in sight. Watch it rebuild the stretch factor.

  1. Lin’s clock ticks twice at her own place, 10 m of time apart. In her frame: \(\dt = 10\), \(\dx = 0\), so \(\tau^2 = 100\). A clock at rest in a frame is the simplest pair of events there is.
  2. Rae’s rocket moves at 0.6, so in Rae’s frame Lin moves at 0.6 the other way. Between the ticks Lin travels \(0.6\,\dtp\). Distance is speed times time, in any frame.
  3. The interval doesn’t care whose notebook it is: \(\dtp^2 - (0.6\,\dtp)^2 = 100\). Since \(0.6^2 = 0.36\), that is \((1 - 0.36)\,\dtp^2 = 100\), so \(0.64\,\dtp^2 = 100\). This is the whole trick. One equation, one unknown.
  4. So \(\dtp^2 = 156.25\) and \(\dtp = 12.5\). Lin moved \(0.6 \times 12.5 = 7.5\). Check: \(12.5^2 - 7.5^2 = 156.25 - 56.25 = 100\). Always check by putting the numbers back in.
  5. Rae measures 12.5 between ticks that Lin’s own clock calls 10. The ratio is \(1.25\): the \(\g\) of Chapter 5, rebuilt from the interval alone. The light clock wasn’t a trick of light. It was the interval all along.

The picture: a circle with a minus sign

The circle on the surveyors’ map The town square at the origin, the Daytimers’ east and north axes in solid blue, and the Nighttimers’ axes in dashed vermilion, turned so their east axis points at the well. A teal circle of radius 500 m passes through the well. The Daytimers say the well is at 300, 400; the Nighttimers say 500, 0. Its rule is east squared plus north squared equals 500 squared. well east north east′ north′ 500 m Daytimers: well at (300, 400) Nighttimers: well at (500, 0) Teal: east² + north² = 500² The hyperbola on the spacetime diagram Event O at the origin of Lin’s spacetime diagram, with gold light lines at 45 degrees. Rae’s worldline, dashed vermilion, runs from O through B at t 5, x 3. A teal hyperbola through B marks every event with t squared minus x squared equal to 16. Lin says B is at 5, 3; Rae says 4, 0. Its rule is t squared minus x squared equals 16. O B t x Rae t² − x² = 16 Lin: B at (5, 3) Rae: B at (4, 0) Teal: t² − x² = 16
Plate 6.1 Same recipe, one sign flipped. With a plus sign, the points 500 m from the square make a circle. With a minus sign, the events with \(t^2 - x^2 = 16\) make a hyperbola. Turning keeps the well on its circle; boosting keeps B on its hyperbola.

Look at the left half first. Every point 500 m from the square sits on one circle, and its rule is east² + north² = 500². Turning the grid changes a landmark’s two numbers, but never which circle it sits on.

Now the right half. Every event with \(t^2 - x^2 = 16\) sits on one curve too. It’s the same recipe: square one number, square the other, combine them. The only change is the sign in the middle, and that one sign changes the shape. Flip it yourself.

One sign, two shapes O at the center, with the gold light lines at 45 degrees. Dashed: t squared plus x squared equals 16, a circle of radius 4 around O. Teal: t squared minus x squared equals 16, a hyperbola with one branch above O and one below, running toward the light lines without touching them. The event (4, 0) is on both; (3.2, 2.4) is on the circle; (5, 3) is on the hyperbola. O (4, 0) (3.2, 2.4) (5, 3) t x Teal: t² − x² = 16 Dashed: t² + x² = 16

With a plus sign, t² + x² = 16 draws a circle of radius 4 around O. With a minus sign, t² − x² = 16 draws a hyperbola: one branch above O and one below, running toward the gold light lines and never touching them. The event (4, 0) is on both.

Flip the sign. Same recipe, same 16, same axes. Only the sign in the middle changes.
Describe this lab

A spacetime diagram with O at the center and the gold light lines at 45 degrees. A two-way switch sets the sign in t² ± x² = 16, shown in large type above the plot. With plus, the curve is a circle of radius 4 around O, drawn in plain ink because t² + x² is not the same in every frame. Switching to minus opens the circle out, through two flat lines, into the teal hyperbola, with one branch above O and one below, which runs toward the light lines and never touches them. The event (4, 0) stays on the curve the whole time; a second marked event moves from (3.2, 2.4) on the circle to (5, 3), Lin’s B, on the hyperbola. A line under the plot does the arithmetic for both events.

Same recipe, one sign flipped: east² + north² = 500² draws a circle, and \(t^2 - x^2 = 16\) draws a hyperbola. The minus sign is the whole difference. It also explains the shape. On a light line, \(t\) and \(x\) are the same size, so \(t^2 - x^2\) is 0 there. A curve where \(t^2 - x^2\) is 16 can never touch a line where it is 0. So it bends away from O and runs alongside the light lines without ever reaching them.

A hyperbola has two halves, or branches: one in O’s future, one in its past. Boosting slides B along its own branch, the bold curve in the boost machine. So the hyperbola is the spacetime circle, drawn with a minus sign. The circle collects every point that is the same distance from the center. The hyperbola collects every event that is the same wristwatch time from O.

Event L is at \((t, x) = (3, 3)\), on O’s light line. B’s curve is \(t^2 - x^2 = 16\). Can a fast enough rocket slide L onto B’s curve?

  • Yes: a boost can slide an event onto any curve.
  • Yes, but only at speeds above 0.6.
  • No: for L, \(t^2 - x^2 = 0\), and every rocket gets 0 too.
  • No, because L is in O’s past.
Hint

Work out \(t^2 - x^2\) for L. Can a boost change that number?

Solution

No. For L, \(t^2 - x^2 = 9 - 9 = 0\). The interval is the same for every rocket, so L stays where \(t'^2 - x'^2 = 0\): on the light line. B’s curve is where the number is 16. That is the minus sign at work: it puts light at 0, so the hyperbola runs toward the light lines but never reaches them.

Why, really?

Where does the minus sign come from? From Pythagoras, used the other way round. In Chapter 5, Lin worked out half a tick of Rae’s light clock as a right triangle. Half of Lin’s time was the slanted side, the longest one: \((t/2)^2 = (\tau/2)^2 + (x/2)^2\). Move one square across the equals sign and you get \(\tau^2 = t^2 - x^2\). The minus sign appears because the wristwatch time is a short side of the triangle, not the long one.

And light keeps it consistent. For a flash, \(x = t\), so \(t^2 - x^2 = 0\), and every observer clocks light at the same speed, so every observer gets that same 0. A plus sign could never give 0 for a flash.

The shorthand

Here is everything above in one line. The colors are the code of the course: blue for Lin’s lab, vermilion for Rae’s rocket, teal for what everyone agrees on.

\[ \inv{\tau^2} = \lab{\dt^2} - \lab{\dx^2} = \rkt{\dtp^2} - \rkt{\dxp^2} \]

Space has three directions, and each one gets a minus sign:

\[ \inv{\tau^2} = \Delta t^2 - \Delta x^2 - \Delta y^2 - \Delta z^2 \]

For example, a firecracker at \((t, x, y) = (13, 5, 12)\) from O gives \(169 - 25 - 144 = 0\). That pair sits exactly on the light cone, though no single coordinate shows it. Need a refresher?

Don’t take one example’s word for it. Here are four notebooks for the same pair O and B. Check the last column yourself.

Four notebooks for the same pair of events O and B
WhoSpeed vΔtΔxΔt² − Δx²
Lin (lab)05325 − 9 = 16 ≡
Rae0.64016 − 0 = 16 ≡
Rocket 20.813/3−5/3169/9 − 25/9 = 16 ≡
Rocket 3−0.68.57.572.25 − 56.25 = 16 ≡
Plate 6.2 Three rockets record one pair. Every notebook disagrees about \(\dt\) and \(\dx\), and every one gets \(\tau^2 = 16\): the last column never changes.

Future, past, elsewhere

A squared number can’t be negative, but \(\tau^2\) can, because it is a difference. Its sign sorts every pair of events into one of three families, and every observer agrees on the family.

Future, past, elsewhere The light cone of event O. Inside the upper cone is O’s future: event A at t 3, x 1, with tau squared 8, reachable by a clock. Inside the lower cone is O’s past: event P at t minus 3, x minus 1, from which a clock can reach O. Along the 45 degree light lines, tau squared is 0: event L at t 4, x minus 4. Outside the cone, left and right, is elsewhere: event E at t 1, x 4, with tau squared minus 15, which no signal from O can reach. O A τ² = 8 P τ² = 8 L τ² = 0 E τ² = −15 light: τ² = 0 FUTURE O can affect these PAST these can affect O ELSEWHERE no signal ELSEWHERE no signal
Plate 6.3 Future, past, elsewhere. The gold light cone of O splits spacetime into three regions, and every observer draws the same cone.
  • Timelike, \(\tau^2 > 0\). Time wins. A clock can ride from one event to the other, and \(\tau\) is what it reads. For A in the plate, \(\tau^2 = 8\): a clock that rides there from O arrives with \(\sqrt 8 \approx 2.8\) on its face.
  • Lightlike, \(\tau^2 = 0\). A tie. Only a flash of light connects the two events. These events make up O’s light cone.
  • Spacelike, \(\tau^2 < 0\). Space wins. Getting from O to E means covering 4 m of space in 1 m of time. Nothing can, not even light. E is elsewhere.

Light leaves a star 4 light-years away and reaches your eye 4 years later. What is the interval between leaving and arriving?

  • Big: the events are light-years apart.
  • Small, but not zero.
  • Zero.
Hint

Use years and light-years: \(\dt = 4\), \(\dx = 4\).

Solution

\(\tau^2 = 4^2 - 4^2 = 0\): the pair is lightlike. Far apart in space and in time, and zero apart in spacetime. The interval is not an ordinary distance.

This is causality in one picture. O can affect only events in its future cone (inside or on it). Only events in its past cone can affect O. Elsewhere is cut off in both directions.

Now the order of events. Take a timelike pair: B at (5, 3). As you boost, B slides along the upper branch of its hyperbola. That half never comes down to \(t' = 0\), so every rocket says B came after O. Now a spacelike pair: B at (3, 5). Its hyperbola lies on its side, and it crosses the line \(t' = 0\). At \(v = 0.6\) the two events are simultaneous, and faster rockets put B first. You saw both in the lab.

Who can be reordered Two pairs, each plotted in the coordinates of three different frames; each dot uses its own frame’s time and space. The timelike event B at t 5, x 3 lands at 4, 0 for a rocket at 0.6 and at 4.33, minus 1.67 for a rocket at 0.8: it rides the top branch of its hyperbola, which never comes down to t = 0. The spacelike event at t 3, x 5 lands at 0, 4 for the rocket at 0.6, the same moment as O, and at minus 1.67, 4.33 for the rocket at 0.8, before O: its hyperbola lies on its side and crosses t = 0. O t, t′ x, x′ (5, 3): lab v = 0.6 v = 0.8 (3, 5): lab v = 0.6: t′ = 0 v = 0.8: t′ < 0 timelike: t′ stays above zero axes: each frame’s own t and x spacelike: slides across t′ = 0
Plate 6.4 Who can be reordered. Think of three frames’ diagrams stacked on one sheet, the lab’s and two rockets’: each dot is the same event, plotted with that frame’s own time and space. The timelike one never reaches \(t' = 0\); the spacelike one slides right across it.

Myth

“If rockets disagree about which event came first, then cause and effect are relative.” No. The order can flip only for spacelike pairs, and a spacelike pair can’t be cause and effect: no signal links them. For every pair that could be cause and effect, timelike or lightlike, every observer agrees which came first.

Proper distance

For a spacelike pair, \(\tau^2\) is negative and no wristwatch can measure it. Flip the sign and you get a positive number: \(\sigma^2 = \dx^2 - \dt^2\). Its square root \(\sigma\) is the proper distance. It is the distance between the events in the one frame where they happen at the same time.

For B at (3, 5): \(\sigma^2 = 25 - 9 = 16\), so \(\sigma = 4\). Your boost machine agreed: at \(v = 0.6\) it put B at \(t' = 0\), \(x' = 4\). Wristwatch time and proper distance are the same invariant, read in the two families where each one makes sense.

Pitfall

“Go fast enough and you’ll reach a spacelike event.” No speed helps. Reaching E from O means 4 m of space in 1 m of time, and every frame agrees \(\tau^2 = -15\). A faster rocket changes the coordinates, never the sign: no boost pulled a card out of the sorter’s elsewhere bin.

Light has no rest frame

One family is odd. Every lightlike pair has \(\tau = 0\). Does that mean light’s own clock stands still?

Light never sits still A light event that the lab places at t 4, x 4, plotted in the coordinates of faster and faster rockets. At v 0.6 it is at 2, 2; at v 0.8 at 1.33, 1.33. Each time it slides along the same 45 degree line toward O, and never reaches the vertical axis x′ = 0 where a light pulse at rest would have to be. t′ x′ O lab: (4, 4) v = 0.6: (2, 2) v = 0.8: (1.33, 1.33) Faster rockets slide it toward O… …but its worldline stays at 45°
Plate 6.5 Light never sits still. A boost slides a light event along its own 45° line. The flash’s whole worldline is that same 45° line in every rocket; sitting still would mean a vertical worldline, along \(t'\).

Myth

“From a photon’s point of view, time stops.” The question has no answer, because a photon has no point of view. To have one, you need a clock riding with you, and clocks can’t reach light’s speed. No boost makes a light pulse sit still either: in Plate 6.5 its worldline stays at 45° in every rocket. So \(\tau = 0\) describes light’s path, not anyone’s experience.

And a zero interval does not mean close. The star’s light covered 4 light-years in 4 years and still has \(\tau = 0\). The interval is not an ordinary distance.

Deep dive: where the rockets’ numbers came from

Plate 6.2 gave each rocket’s numbers without saying how. Here is a way to get them using nothing beyond multiplying, ahead of the full machine in Chapter 10.

Use two new coordinates, the sum \(s = t + x\) and the difference \(d = t - x\). Their product is the interval: \(s\,d = (t + x)(t - x) = t^2 - x^2\). For B at (5, 3): \(s = 8\), \(d = 2\), and \(8 \times 2 = 16\).

A boost at \(v = 0.6\) turns out to halve \(s\) and double \(d\). So Rae gets \(s' = 4\), \(d' = 4\). Back to \(t\) and \(x\): \(t' = (s' + d')/2 = 4\) and \(x' = (s' - d')/2 = 0\). That’s Rae’s (4, 0). Halving one number and doubling the other can’t change their product, which is why \(\tau^2\) survives.

At \(v = 0.8\) the factor is 3 instead of 2: \(s' = 8/3\), \(d' = 6\), so \(t' = 13/3\) and \(x' = -5/3\). At \(v = -0.6\) the roles swap: \(s' = 16\), \(d' = 1\), so \(t' = 8.5\) and \(x' = 7.5\). In general the factor is \(\sqrt{(1+v)/(1-v)}\). It will come back in Chapter 12 as the Doppler factor, the number that sets the color of light from a moving source.

That’s the physics of the chapter: one number, three families, and a guarantee about cause and effect. Next question: has Nature ever broken the guarantee?

Ask Nature

If anything ever crossed from O into its elsewhere, the guarantee would fail. A message could reach a rocket that measures it arriving before it was sent. So physicists keep looking for anything faster than light. In more than a century, nothing has ever been caught outside the light cone. Once, it looked as if something had.

Ask Nature

In 2011 the OPERA experiment fired neutrinos (tiny particles that pass through rock almost untouched) from CERN, near Geneva, through the Earth to a detector under the Gran Sasso mountain in Italy. The neutrinos seemed to arrive about 60 ns sooner than light would have. Their worldline would have been just outside the light cone. OPERA Collaboration, Journal of High Energy Physics 10 (2012) 093 (arXiv:1109.4897)

The team didn’t announce that relativity was dead. They published their numbers and asked everyone to help find the mistake. In 2012 they found it: the main culprit was a loose fiber-optic connector in the timing system. With it fixed, the neutrinos arrived on time.

Real numbers

How big an error is 60 ns? One nanosecond is about 0.2998 m of time, so \(60 \times 0.2998 \approx 18\) m of time. The neutrinos seemed to beat a light flash by the time light needs to cross a bowling lane, after a trip through hundreds of kilometers of rock.

Light’s side of the bargain has been tested as well, by the CERN experiment you met in Chapter 2. In 1964, photons from particles that were themselves moving at about 0.99975 of light speed were timed in flight. They still moved at light speed, to within about one part in 10 000. A source that is moving doesn’t tilt its own light cone. That’s why everyone can draw the same cone.

Practice

Three tiers: ◇ back of the envelope, ◆ pencil and paper, ★ head-scratcher. Units are meters of time and meters unless the problem says years and light-years.

Events A = (0, 0) and B = (5, 3). A clock rides straight from A to B. What does it read when it gets there?

Hint

Square each separation, then subtract space from time.

Hint

\(5^2 - 3^2 = 25 - 9\). The clock shows the square root.

Solution

\(\tau^2 = 5^2 - 3^2 = 16\), so \(\tau = 4\). It’s Rae’s clock reading from the puzzle.

Event A happens on Earth. Event B happens 3 years later, on a planet 5 light-years away. Could A have caused B?

  • Yes: B comes after A, so there was time.
  • Yes, with a fast enough signal.
  • No: the pair is spacelike.
  • It depends on who is watching.
Hint

Compute \(\tau^2\) in years: which is bigger, \(\dt^2\) or \(\dx^2\)?

Solution

\(\tau^2 = 3^2 - 5^2 = -16 < 0\): spacelike. A signal would need to cover 5 light-years in 3 years, faster than light. So A could not have caused B, in any frame.

For the same pair (\(\dt = 3\) years, \(\dx = 5\) light-years), find the proper distance \(\sigma\).

Hint

\(\sigma^2 = \dx^2 - \dt^2\).

Solution

\(\sigma^2 = 25 - 9 = 16\), so \(\sigma = 4\) light-years. A rocket at \(v = 0.6\) measures the two events as simultaneous, 4 light-years apart.

Lin’s clock ticks twice at the same place: \(\dx = 0\) and \(\dt = 10\). A passing rocket measures the two ticks \(\dxp = 8\) apart. How much time passes between them in the rocket’s frame?

Hint

In the lab, \(\tau^2 = 10^2 - 0^2 = 100\).

Hint

In the rocket, \(\dtp^2 - 8^2\) must also be 100.

Solution

\(\dtp^2 = 100 + 64 = 164\), so \(\dtp = \sqrt{164} \approx 12.81\). It’s longer than 10: Lin’s clock is the one present at both ticks, and it reads the least time of anyone.

Each card gives the separation \((\dt, \dx)\) of a pair of events. Sort them.

  • (5, 3)
  • (3, 5)
  • (4, 4)
  • (13, 5)
  • (1, 0)
  • (0, 2)
Hint

Only the sign of \(\dt^2 - \dx^2\) matters. Which is bigger, the time part or the space part?

Solution

\(\tau^2\) is 16, −16, 0, 144, 1 and −4. Timelike: (5, 3), (13, 5), (1, 0). Lightlike: (4, 4). Spacelike: (3, 5), (0, 2). Note (0, 2): two events at the same moment are always spacelike, however close.

Place an event B that is lightlike-separated from A = (0, 0), in A’s future, and to the left of A.

No pointer? Type B’s t and x into the boxes under the diagram.

Hint

Lightlike means \(|\dx| = |\dt|\): B sits on a 45° line through A.

Hint

Future means \(t > 0\). Left means \(x < 0\). So \(x = -t\).

Solution

Any point on the line \(x = -t\) with \(t > 0\), for example \((3, -3)\). A flash from A heading left passes through it.

A pair of events is timelike: a clock can ride from the first to the second. Can a fast enough rocket measure the second one happening first?

  • Yes, if the rocket goes fast enough.
  • Yes, but only above v = 0.6.
  • No: every frame agrees on the order of a timelike pair.
  • Only light could reverse it.
Hint

A boost slides B along its hyperbola. For B at \((5, 3)\), does the upper branch of \(t^2 - x^2 = 16\) ever come down to \(t' = 0\)?

Solution

No. A boost slides the later event along its branch of the hyperbola, and for a timelike pair that branch stays entirely in the future of the first event. Only spacelike pairs can be reordered.

Finish this worked example. Rae’s wristwatch ticks off 6 m of time between two events at her own place. Her rocket passes Lin at \(v = 0.8\). How far apart are the events in Lin’s frame, in time and in space?

  1. In Rae’s frame, \(\dtp = 6\) and \(\dxp = 0\), so \(\tau^2 = 36\).
  2. In Lin’s frame Rae moves at 0.8, so \(\dx = 0.8\,\dt\).
  3. Invariance: \(\dt^2 - (0.8\,\dt)^2 = 36\), so \(0.36\,\dt^2 = 36\).
  4. Your turn: \(\dt = \) ? and \(\dx = \) ?

Hint

\(0.36\,\dt^2 = 36\) gives \(\dt^2 = 100\).

Solution

\(\dt^2 = 36 / 0.36 = 100\), so \(\dt = 10\), and \(\dx = 0.8 \times 10 = 8\). Check: \(100 - 64 = 36\). The ratio \(10/6 = 5/3\) is \(\g\) at 0.8.

In three dimensions: A = (t 0, x 0, y 0) and B = (t 7, x 2, y 3). What does a clock read after riding straight from A to B?

Hint

Every space direction gets a minus sign: \(\tau^2 = \dt^2 - \dx^2 - \Delta y^2\).

Solution

\(\tau^2 = 49 - 4 - 9 = 36\), so \(\tau = 6\).

Lin and Rae (rocket at 0.6) both record the same two events. Tick everything they are guaranteed to agree on.

  • The squared interval \(\tau^2\)
  • The time between the events
  • The distance between the events
  • Whether the pair is timelike, lightlike or spacelike
  • Which event came first, if the pair is timelike
  • Which event came first, if the pair is spacelike
  • The speed of a light pulse
Hint

Check each item against the two notebooks for O and B: \((5, 3)\) and \((4, 0)\). Anything that differs between them is not guaranteed.

Solution

Guaranteed: \(\tau^2\), and with it the family; the order of a timelike pair; light’s speed. Not guaranteed: \(\dt\), \(\dx\), and the order of a spacelike pair.

A ship leaves Earth at \(t = 0\) and coasts at constant speed to a star 6 light-years away, arriving at \(t = 10\) years in Earth’s frame. How many years pass on the ship’s own clock?

Hint

The ship’s clock is present at both events, so it reads \(\tau\).

Solution

\(\tau = \sqrt{10^2 - 6^2} = \sqrt{64} = 8\) years. (The ship moves at \(6/10 = 0.6\), and \(10 / 8 = 1.25 = \g\).)

Two events have \(\tau^2 = -9\). In Lin’s frame they are \(\dt = 4\) apart in time. How far apart are they in Lin’s frame? And how far apart in the frame where they happen at the same moment?

Hint

\(4^2 - \dx^2 = -9\).

Hint

Where they are simultaneous, \(\dtp = 0\), so \(-\dxp^2 = -9\).

Solution

\(\dx^2 = 16 + 9 = 25\), so \(\dx = 5\). In the simultaneous frame \(\sigma^2 = 9\), so \(\sigma = 3\). Notice the proper distance is the smallest distance any frame measures between them.

For which separation \((\dt, \dx)\) can some rocket measure both events at the same place?

  • (2, 6)
  • (6, 2)
  • (4, 4)
  • Any of them, with the right rocket
Hint

“Same place” means \(\dxp = 0\). Then \(\tau^2 = \dtp^2\), which must be positive.

Solution

Only a timelike pair: (6, 2), with \(\tau^2 = 32\). A rocket at \(v = 1/3\) is present at both. Timelike pairs have a same-place frame; spacelike pairs have a same-time frame; lightlike pairs have neither.

Stuck on a square root? The math toolkit has a refresher, and the formula card has this chapter’s equations on one page.

Build it

Three small functions, and the labs above start running on your code. Once they pass, the boost machine’s τ² column and its timelike, lightlike and spacelike tags, and the sorter’s verdicts, all come from you.

Code you need

Comparisons and choices. >, <, >= and <= compare two numbers. if runs a line only when its test is true; else catches the rest. A function can return words, written in quotes:

function weather(celsius) {
  if (celsius > 25) return 'hot';
  else if (celsius < 5) return 'cold';
  else return 'mild';
}

Missing values. An event may have no y. Then e.y is undefined, and e.y ?? 0 means “e.y, or 0 if it’s missing”:

const pool = { width: 10 };
// pool has no depth: gives 2
const depth = pool.depth ?? 2;

Not a number. Math.sqrt(-4) gives NaN, short for “not a number”. You can also return NaN on purpose, to say “this has no answer”.

Size without sign. Math.abs(-3) is 3: the size of a number, without its sign. Math.abs(3) is 3 too.

Build it: intervalSquared(a, b)

Return \(\dt^2 - \dx^2 - \Delta y^2 - \Delta z^2\) for events a and b. Events look like {t: 5, x: 3}. One test gives them a y as well, so treat a missing y or z as 0 with ??. The tests also swap the two events: the answer must not change.

Build it: classify(a, b, eps)

Return 'timelike', 'lightlike' or 'spacelike'. One catch: computers round. A pair that should give \(\tau^2 = 0\) can come out as 0.000000000002. So count \(\tau^2\) as zero when it is tiny compared with the size of the separation: when \(|\tau^2| \le\) eps \(\times (\dt^2 + \dx^2 + \Delta y^2 + \Delta z^2)\). The third input, eps, is already in the starter with a default of 1e-9, so the tests can leave it out.

You can call intervalSquared(a, b) here: it is your passing version (or the reference one, until yours passes). You still need dt, dx, dy and dz for the rounding band, so copy those lines from intervalSquared too.

The tests check (5, 3), (3, 5) and (4, 4); a rounding error of \(10^{-12}\); a pair with a y, (13, 5, 12); and two firecrackers at the same moment, 10 μm apart, which must come out spacelike. Tiny is not the same as zero.

Build it: properTime(a, b)

The wristwatch time of a clock riding straight between a and b: \(\sqrt{\tau^2}\). For a spacelike pair no clock can make the trip, so return NaN. A light ray’s pair gives 0. The tests also list the later event first: answer with the same \(\sqrt{\tau^2}\) either way, because \(\tau^2\) doesn’t care about the order. As with classify, you can call intervalSquared(a, b) here.

Checkpoint

Five graded items, plus one written answer at the end. Feedback waits until you submit.

A star 8 light-years away flares. The flash reaches Earth 8 years later. What is \(\tau^2\) between the flare and the flash’s arrival? (Use years and light-years.)

Solution

\(\tau^2 = 8^2 - 8^2 = 0\). The events are 8 light-years and 8 years apart, and the interval between them is zero: they are lightlike.

Event O is at the origin. Each card gives another event’s \((t, x)\) from O. What is possible?

  • (3, 1)
  • (−4, 2)
  • (1, 5)
  • (−2, −2)
  • (2, −2)
Solution

(3, 1) and (2, −2) are in O’s future cone (the second on it: a flash reaches it). (−4, 2) and (−2, −2) are in O’s past cone. (1, 5) is spacelike: neither can affect the other.

Place an event P that happens after O in Lin’s frame, but that some rocket measures happening before O.

No pointer? Type P’s t and x into the boxes under the diagram.

Solution

Any spacelike event with \(t > 0\): \(|x| > t\), for example \((1, 4)\). Its hyperbola crosses \(t' = 0\), so a fast enough rocket puts it before O. A timelike event, like \((4, 1)\), can never be reordered.

Sagredo: “Light’s interval is zero. So if I rode along with a light pulse, my watch would stand still.” What’s wrong with that?

  • Nothing: from a photon’s point of view, time stops.
  • No frame rides with light: no rocket slower than light makes a pulse sit still, and no clock can reach light speed.
  • The watch would run backward instead.
  • Only the hands stop; the rest of time goes on.
Solution

Riding along with light would need a frame where the pulse is at rest. Every frame clocks light at speed 1, and boosting only slides a light event along its 45° line. So the question has no answer: light has no rest frame.

Event E is 2 m of time after O and 5 m away. Sagredo can’t get from O to E at \(v = 0.9\). “At 0.99 I’ll make it,” he says. Will he?

  • Yes, at 0.99.
  • Yes, at 2.5 times light speed.
  • No: E is spacelike from O, so no speed gets him there.
  • It depends on which frame describes the trip.
Solution

\(\tau^2 = 2^2 - 5^2 = -21\): spacelike, in every frame. A traveler from O to E would cover 5 m of space in 2 m of time, faster than light. Speed changes coordinates, never the sign of \(\tau^2\).

In two or three sentences: why can a rocket reverse the time order of O and a spacelike event, but never of O and a timelike event?

Model answer

A boost slides each event along its own hyperbola of fixed \(\tau^2\). A timelike event rides one branch of its hyperbola, which stays on one side of O in time, so its order never changes. A spacelike event’s hyperbola lies on its side and crosses \(t' = 0\), so some rockets measure it as simultaneous with O, and faster ones put it first. That’s harmless, because nothing can travel between a spacelike pair.

  • Boosts keep each event on its hyperbola
  • A timelike event’s branch never crosses t′ = 0; a spacelike one does
  • Spacelike pairs can’t be cause and effect, so causality is safe

Explain it to Sagredo

Sagredo: “On the surveyors’ map, the points 500 m from the square make a circle. You say one minus sign turns that circle into a hyperbola. Why, and why should I care?”

Model answer

The circle is the set of points with the same distance from the square, \(x^2 + y^2 = 500^2\), and turning the map only moves points around it. In spacetime the set of events with the same interval from O is \(t^2 - x^2 = \text{constant}\), and with a minus sign that equation draws a hyperbola instead of a circle. Boosting slides events along their hyperbolas, just as turning slides landmarks around their circles. You should care because that is what keeps cause and effect safe: a timelike event in O’s future rides the future branch of its hyperbola, which never crosses into O’s past.

  • The curve is a set of equal-interval events
  • t² − x² = constant is a hyperbola
  • Boosts slide events along it

What’s invariant here?

  • The interval. \(\inv{\tau^2} = \dt^2 - \dx^2 - \Delta y^2 - \Delta z^2\) is the same in every free-float frame. For a timelike pair, \(\tau\) is the wristwatch time of a clock riding straight between the events.
  • Its sign, so the family. Timelike (\(\tau^2 > 0\)), lightlike (\(= 0\)), spacelike (\(< 0\), with proper distance \(\sigma = \sqrt{-\tau^2}\)).
  • Time order, where it matters. Every frame agrees on the order of a timelike or lightlike pair. Only spacelike pairs can be reordered, and nothing travels between them.
  • The light cone. Boosts slide events along hyperbolas, the way turning slides points around circles. Light’s 45° lines never move.

Still strange?

Two firecrackers can go off in one order for Lin and the other order for a fast rocket, and nobody is wrong. Nature protects cause and effect with a single minus sign; we know very precisely that it works, and nobody knows why the world had to be built that way.

Going further

Use the deep dive’s trick. Lin records an event at \((t, x) = (10, 6)\). A rocket at \(v = 0.6\) halves \(s = t + x\) and doubles \(d = t - x\). Where does the rocket put the event?

Hint

\(s = 16\) and \(d = 4\), so \(s' = 8\) and \(d' = 8\).

Solution

\(t' = (s' + d')/2 = 8\) and \(x' = (s' - d')/2 = 0\). The event sits on the rocket’s own worldline, 8 m of time along it. Check: \(10^2 - 6^2 = 64 = 8^2\).

A rumored particle, the tachyon, would always travel faster than light. Which family would every pair of events on its worldline belong to?

  • Timelike
  • Lightlike
  • Spacelike
Hint

Between two events on its worldline, which is bigger: the space it covers or the time it takes?

Solution

Spacelike: it covers more space than time. That is exactly the family whose order rockets disagree about, which is why tachyons would wreck cause and effect. None has ever been found.

Reading. Taylor and Wheeler’s Spacetime Physics builds the whole subject on the interval, and the surveyors’ parable comes from it; see eftaylor.com. Feynman’s chapter on space-time in volume I of The Feynman Lectures on Physics treats the same idea with rotations, free to read at feynmanlectures.caltech.edu.