物理 · Physics

Special Relativity

Day 14 · 2026 · Phase D Relativity
What gets relativised isn't the laws of physics but the word "simultaneous" — to keep one speed invariant, spacetime gives up having a universe-wide "now".
The name has been misleading readers for a century. The theory's claim is not that "everything is relative" — quite the opposite: it insists that some things are absolutely invariant, namely the laws of physics and that speed called c. The trouble is that once you demand an invariant speed, time and length have to give way. And the first casualty is the concept you'd have bet on last: two events happening "at the same time".

The Invariance of c

principle of relativity · 1905
Intuition At sixteen Einstein asked a childlike question: if I chase a light beam at the speed of light, what do I see? By everyday experience (catch up with a car at its own speed and it stands still to you), you ought to see a beam of frozen light. But no such thing exists in Maxwell's equations — they only allow waves travelling at one fixed speed, and the c = 1/√(ε0μ0) they hand you (those subscripts are the digit 0, zero, meaning vacuum; see "The Unification of Electromagnetism") has no slot at all for "relative to whom".
Mechanism Einstein's move was to accept that as a fact and see what it costs. Two postulates: ① the principle of relativity — the laws of physics take the same form in every inertial frame (a frame moving uniformly in a straight line, not accelerating); ② the invariance of light speed — the vacuum speed of light is the same c for every inertial observer, independent of how the source or the observer moves. The cost is that velocities no longer simply add:
w = u + v1 + uv/c²
A train moves at v, someone aboard throws a ball forward at u (relative to the train), and the ground sees speed w. The uv/c² in the denominator is the whole of what's new: at everyday speeds it is absurdly tiny and the formula collapses back to the schoolroom w = u+v; but put u = c and the answer is always still c. Light isn't a speed you "can't add to" — it's this rule's fixed point.
The counterintuitive part c is not "the speed of light"; light merely happens to reach it. c is spacetime's own conversion rate — the exchange rate turning "one second" into "this much distance". Anything with no rest mass travels at it (light, gravitational waves); anything with mass never gets there. That step promotes c from a property of some wave to a structural constant of spacetime — and since 1983 the metre has simply been defined as the distance light covers in 1/299792458 of a second: the speed of light is no longer measured, it's fixed by decree.
Cross-disciplinary · chips / networks / deep space A finite c is a wall engineering runs into daily:
  • Chips: at 5 GHz one clock period is 0.2 nanoseconds, in which light travels 6 centimetres — shorter than the motherboard. A signal crossing the die misses the beat, which is why clock skew and multiple clock domains exist: a single chip no longer has one unified "now" inside it.
  • Data centres: two machines 100 metres apart need at least 0.67 microseconds for a round trip; across the Atlantic, at least 60 milliseconds. That's a physical floor, and every difficulty in "globally strongly consistent databases" follows from it.
  • Deep space: Earth–Mars one-way comms take 3–22 minutes, so a rover must decide for itself. "Remote driving" is physically off the table.
In a sentence: c isn't the speed of light, it's spacetime's conversion rate.
Ponder: If nothing can outrun c, why do astronomers say some galaxies recede from us faster than light?
Because that isn't "running through space" — it's space itself getting bigger. Relativity limits motion through space and says nothing about the expansion of space. Nobody is chasing anybody; new space keeps being made between them. It breaks no rule, and it can't carry a signal.

Relativity of Simultaneity

simultaneity · the train experiment
Intuition One carriage, a lamp hung at its exact centre, a detector at each end. The passenger: equal distances, equal speed, both ends light up at once. The observer on the platform: the train is moving forward, so the rear wall runs into the leftward light while the front wall flees from the rightward light — and light travels at c for him too (that's the postulate, no discounts), so the light reaches the rear wall first. Neither has miscalculated, and neither may call the other wrong: the word "simultaneous" simply doesn't refer to the same thing in their two mouths.
On the train: both ends light up at once rear ✓ lit front ✓ lit equal paths → arrive together On the ground: the rear wall lights first where the lamp flashed (ground coordinates) light: 100 light: 100 train speed v rear runs into it ✓ lit front is fleeing ✗ not yet
Both beams cover the same distance (light is c for everyone), but the walls themselves are moving: the rear one runs into it, the front one ducks away.
Mechanism Written out, this is the time line of the Lorentz transformation:
Δt′ = γ ( Δtv Δxc² )
Δ ("delta") means "the difference between two events": Δt is the time gap measured in one frame and Δx their spatial separation, Δt′ ("t prime") is the time gap in the other frame, v is the relative speed of the frames, and γ ("gamma") is for now just a number slightly above 1. The crux is the second term vΔx/c²: it is non-zero only when Δx is non-zero, i.e. when the events happen in different places. Two events at the same spot are simultaneous for everybody; the further apart they are, the wider the disagreement about "at once".
The counterintuitive part There is no global snapshot called "now". "What is happening in the Andromeda galaxy at this moment" has no unique answer: the "now" slice you cut while strolling and the one you cut standing still are days apart out at 2.5 million light years.
But don't slide from there into mysticism: this does not mean your walking influences Andromeda. Those events are spacelike separated from you — unseeable, untouchable, 2.5 million years before light could link you. What relativity actually guarantees is that for any two events with a causal connection, every observer agrees on the order; the ones whose order can be shuffled are precisely those that can never reach each other, where the order makes no difference anyway.
Cross-disciplinary · distributed systems Computer science hit the same wall and came out with almost the same structure:
  • Partial order instead of a timeline: Lamport's 1978 classic points out that a distributed system has no global clock — events only stand in a partial order called happens-before. Those that can influence each other have a definite order; those that can't are "concurrent" and may be sorted arbitrarily. Same diagram as "causal order absolute, spacelike order observer-dependent". Git recording commits as a DAG (directed acyclic graph) rather than a timeline is exactly this partial order.
  • Putting the uncertainty in the interface: Google Spanner's TrueTime returns not an instant but an interval [earliest, latest], and a transaction waits out that interval before committing — an admission that simultaneity is undecidable and can only be bought back with waiting. Stock exchanges go the other way, paying dearly for a single matching engine that manufactures an absolute "now".
In a sentence: the universe has no shared "now", only the slice each observer cuts.
Ponder: If observers can disagree about which event came first, why doesn't an effect ever precede its cause?
Because the disagreement only arises between spacelike events — too far apart in space and too close in time for even light to make the trip, so no causal link was possible in the first place. For any two events that a not-faster-than-light signal can connect, every inertial frame computes the same order. Relativity traded the absoluteness of simultaneity for the absoluteness of causation.

Time Dilation & Length Contraction

light clock · the γ factor
Intuition Build the most honest clock there is: two parallel mirrors with a light pulse bouncing between them, up and back counting as one "tick". At rest the light runs a vertical path; set the whole clock flying sideways and, from where you stand, the light is carried forward while it climbs, so it runs a slanted path — and the hypotenuse is longer. Yet light still moves at c for you, not a hair faster. Longer path, same speed: one tick has to take more time. A moving clock runs slow not because the clock is faulty, but because time is.
Clock at rest (its own frame) light: c·t₀ one tick = t₀ (proper time) Same clock flying past (your frame) c·t/2 c·t₀/2 v·t/2 v longer hypotenuse → one tick = t > t₀
One clock, one tick: straight edges on the left, slanted on the right. Light speed is fixed, so the slanted trip costs more time — the right triangle hands you γ directly.
Mechanism One use of Pythagoras on the right-hand figure does it. Over half a tick the hypotenuse is the light's path ct/2, the vertical leg is the mirror gap ct0/2, the horizontal leg is how far the clock itself travelled, vt/2. So (ct)² = (ct0)² + (vt:
t = γ t0 , γ = 1√(1 − v²/c²)
t0 (that subscript is the digit 0, zero) is the time the clock measures for itself, called proper time; t is the same tick as measured by you standing beside it. γ ("gamma") is always ≥ 1: at v = 0 it equals 1 (no difference at all), and the closer v gets to c, the closer the square root in the denominator comes to the digit 0 (zero), sending γ to infinity. The same γ also governs length contraction: a moving object is 1/γ as long along its direction of travel.
A few numbers: at 0.1c, γ ≈ 1.005; at 0.5c, 1.15; at 0.87c, exactly 2; at 0.99c, 7.1. A jet's γ exceeds 1 by about 5×10⁻¹³ — a lifetime of flying won't accumulate a millisecond. Relativity doesn't switch off at low speed; it's just too small to catch.
1 2 8 γ factor 0 0.5c c speed v the v = c wall nothing with mass gets here γ=2 takes 0.87c everyday speeds all sit in here, γ≈1
The curve hugs 1 across most of the range — which is why nobody noticed relativity for two thousand years; only near c does it rear up and diverge.
The counterintuitive part The sky verifies this every second: cosmic rays create muons (particles like electrons but 200 times heavier) about 15 km up, and a muon lives on average 2.2 microseconds — enough for 660 metres even at light speed, so it should never reach the ground. Yet detectors collect them by the minute. The beauty is that the two frames give completely different accounts that agree: from the ground, the muon's clock is slowed by γ ≈ 20 and it lives long enough; from the muon, its lifetime is untouched and the 15 km of atmosphere has been squeezed to a few hundred metres. Time dilation and length contraction aren't two effects; they're two slices of one thing.
Cross-disciplinary · navigation / accelerators
  • Satellite navigation: a GPS satellite's atomic clock runs about 7 microseconds a day slow because of its speed (that's only the special-relativistic share; the gravitational share points the other way and is larger — that's for the "General Relativity" issue). Uncorrected, the positional error piles up to kilometres per day: the blue dot on your phone casts a signed vote for relativity daily.
  • Particle accelerators: protons in the LHC run at γ ≈ 7500, so the 27-kilometre ring they "feel" is 3.6 metres long. Magnet timing, beam lifetime and collision energy are all computed relativistically.
In a sentence: a moving clock isn't wrong — the path it takes is a different length.
Ponder: Motion is relative, so A sees B's clock run slow and B sees A's do the same. Isn't that a contradiction?
No. Comparing clocks in two different places requires first agreeing what "simultaneous" means, and their planes of simultaneity are tilted with respect to each other. As long as they never meet, no objective fact decides between them and both accounts are self-consistent. Once they do meet (the twins), one of them must have switched frames along the way, the symmetry breaks, and the answer is unique.

What E=mc² Really Says

mass–energy equivalence · 1905
Intuition The popular reading is "mass can turn into energy", as though mass were a fuel you burn for energy. The accurate reading is that mass is already a form of energy — another name for "the part of the energy that is sitting still". c² is no mysterious conversion engine, just an exchange rate between units (kilograms into joules), like the 3600 in "1 hour = 3600 seconds".
Mechanism The full expression isn't the famous line but this one:
E² = (pc)² + (mc²)²
E is total energy, p is momentum (how hard it's moving), m is the rest mass (also called invariant mass — the value every observer agrees on). Two limits are the useful ones: for an object at rest, p = the digit 0 (zero) and the formula collapses to E = mc², so the famous line is only the at-rest special case; a photon has no rest mass, m = 0 (zero), and the formula becomes E = pc — no mass, yet energy and momentum, which is what pushes a solar sail.
The structure is itself a right triangle: mc² is one leg (what you have standing still), pc is the other (what you get by moving), and the total energy E is the hypotenuse. However you run, the mc² leg keeps its length — hence the name "invariant mass".
The counterintuitive part A nuclear reaction does not annihilate mass to buy energy. What happens is that the fragments end up bound more tightly, the binding energy changes, and so they weigh slightly less than what went in, the difference having escaped as kinetic energy and radiation. Seal all that escaping energy inside a closed box and weigh it, and the total mass is unchanged to the last decimal. Conservation of mass and conservation of energy were never two laws; they are one.
A harder blow: about 99% of your body's mass is not the rest mass of its quarks but the strong-interaction field energy binding those quarks inside protons; what the Higgs mechanism supplies directly is roughly 1% (the issue on the origin of mass and symmetry will do that accounting). "Mass is energy" isn't a literary flourish — it's the literal composition of your body weight.
Cross-disciplinary · astrophysics / energy / chemistry
  • Stars and life: the Sun turns about 4×10⁹ kilograms (four million tonnes) of mass into radiation every second. Photosynthesis, food, fossil fuels, your body heat right now — all downstream of that one exchange machine.
  • Energy: a 1-gigawatt nuclear plant running flat out for a year truly "loses" about one kilogram of mass. Conversely, one gram is about 9×10¹³ joules — the Hiroshima bomb released roughly 6×10¹³ joules, a mass deficit of under a gram.
  • Chemistry: chemical reactions have a mass deficit too, about a hundred million times smaller — burn a tonne of coal and the products are about 0.3 micrograms lighter, undetectable on any balance, which is why the "conservation of mass" taught in chemistry holds perfectly at its own precision.
In a sentence: mass isn't the raw material of energy — mass is energy locked in place.
Ponder: Heat a sealed box of gas so the molecules move faster. Does the box get heavier?
Yes. A box's invariant mass includes every kind of energy inside it divided by c², molecular kinetic energy included — about 1.1×10⁻¹⁷ kg per extra joule. Unmeasurable, but unambiguous in principle: a wound spring outweighs a slack one, hot coffee outweighs cold.

Going Deeper

Why is "relativity" a badly chosen name?
Because it suggests the subject is what varies, when the skeleton of the theory is what doesn't. Besides light speed there is a deeper invariant: the spacetime interval between two events, s² = (cΔt)² − Δx². Observers disagree about Δt and Δx separately, yet that combination comes out the same for everyone — like a stick whose projected length changes with viewing angle while the stick's own length does not. Einstein at one point favoured calling it "the theory of invariants". That invariant is the front door to the issue on spacetime.
The twin paradox: who really ends up younger, and on what grounds?
The one who flies out and comes back, uncontroversially. The resolution is that the two are not symmetric: the stay-at-home sits in one inertial frame throughout, while the traveller must turn around, and turning around means switching frames. The prettier statement comes from spacetime geometry: each person's proper time is the "length" of their own worldline, and in spacetime the straighter the path, the more proper time it holds (exactly the reverse of "a straight line is the shortest path" in Euclidean geometry). The stay-at-home goes straight, so they age the most.
Lorentz and Poincaré wrote down the same transformation first. Why does the credit go to Einstein?
The formulae did come first. Lorentz read length contraction as a real deformation of an object moving through the ether, and his "local time" was to him a mathematical convenience; Poincaré came closer still but wouldn't part with the ether either. Einstein's step was ontological: he threw the ether out entirely and declared these to be not the dynamics of objects but the kinematics of spacetime itself — every inertial frame's time equally real, none more "true" than another.

Further Reading