Thought Toys · Waves & rhythm · Exhibit 70

The angle past which light refuses to leave

A beam of light inside glass, aimed at the surface. Tilt it and the escaping ray bends further and further over, the reflection underneath growing brighter — until one exact angle. Past it, the escaping ray doesn't dim. It stops existing, and the surface becomes a perfect mirror. Every fiber-optic cable on Earth is built on the far side of that angle.

The surface, and how much light crosses it the beam overall s-polarised p-polarised

Try it
your turn — drag the tilt slowly through 41.8° and watch the escaping ray die

What you're seeing

A side view of the surface between glass (below) and air (above), with a beam of light inside the glass aimed up at it. At every tilt the beam splits in two: a refracted ray that escapes into the air, and a reflected ray that bounces back down. The brightness of each ray on screen is exactly its share of the light — no dramatisation. Straight on, glass lets 96% through and reflects 4%, which is why a window at night is a faint mirror. The graph on the right tracks the reflected share across every tilt at once. (The two thin curves are the two polarisations — the two directions the light wave's vibration can point. The green one falls to zero around 34°: at that tilt, Brewster's angle, glass simply cannot reflect that component. Polarised sunglasses exploit exactly this to erase glare.)

Now drag the tilt up and watch two things at once. The escaping ray bends further and further away from the normal — always more tilted than the beam inside, because light entering thinner stuff swings outward. And the reflected share climbs: slowly at first, then steeply. At 41.8° the escaping ray lies flat along the surface itself. One half-degree more and there is no direction left for it to go — refraction would need the sine of an angle to exceed 1, and no angle has one. The light doesn't mostly reflect. It all reflects: exactly 100%, to the last photon. No silvering, no coating — a perfect mirror made of pure geometry.

That perfection is the trick behind fiber optics. A glass thread carries light for kilometers by letting it bounce off the walls thousands of times per meter — survivable only because each bounce past the critical angle loses nothing at all. An ordinary household mirror eating 5% per bounce would go black in centimeters. Diamond plays the same trick on your eye: press Diamond, 30° and see how low its wall sits — 24.4°. Almost any ray inside a cut diamond is past it, so light entering the stone rattles around the facets before finding a way out. The sparkle is trapped light, escaping on the jeweller's schedule.

Then press send the light in from outside. Same surface, same physics, opposite direction — and the wall is gone. Going into denser stuff, light bends toward the normal, so no entry tilt — not even 89.5°, skimming the surface — can be refused. The trap is strictly one-way: easy to enter, and past one angle, impossible to leave.

The rule, exactly. Crossing between media, the tilt obeys Snell's law n₁ sin θ= n₂ sin θ Leaving a dense medium (n₁ > n₂), sin θ₂ = (n₁/n₂) sin θ₁ reaches 1 at the critical angle θc = asin(n₂/n₁) — 41.8° for glass to air, 48.8° for water, 24.4° for diamond. Past it no real θ₂ exists and reflection is total. Below it, the reflected share of each polarisation comes from the Fresnel equations, e.g. Rs = ((n₁cosθ₁ − n₂cosθ₂)/(n₁cosθ₁ + n₂cosθ₂))² and the beam's brightness on screen is the average of the two. Verified in node (improve/verify/70-total-internal-reflection.js): energy is conserved (reflected + transmitted = 1) to 10⁻¹² across 36,214 cases and all 50,760 slider-reachable states; the critical angle is sharp (T > 0 just below, R = 1 exactly at and above); Brewster's angle falls out of the same formulas unasked (Rp < 10⁻²⁰ at atan(n₂/n₁)); and going into the denser medium never traps at any tilt. Three negative controls fail as they must: the classic amplitude-vs-energy mistake (using t² as the transmitted share) breaks conservation by 379%; the wrong-formula critical angle acos(n₂/n₁) claims 45° glass-to-air light escapes when Snell would need sin θ₂ = 1.06; and a hardcoded 45° wall misclassifies both water (48.8°) and glass (41.8°). Past the wall a faint evanescent skin of field does extend into the air a few hundred nanometers deep — but it carries no energy away, and the total reflection stands.

Also in Waves & rhythm: Fourier epicycles →

All 7 in Waves & rhythm
  1. 12Fourier epicycles
  2. 19Standing waves & normal modes
  3. 21Lissajous figures
  4. 62Ask which slit, lose the wave
  5. 64Outrun your own sound
  6. 66Move, and your clock falls behind
  7. 70The angle past which light refuses to leave — you are here

← the cabinet · Thought Toys — a cabinet of explorable explanations. Exhibit 70.