⭐ The Bubble That Makes Light — an interactive tour

ProtoDynamics · sonoluminescence, live-simulated in your browser — and the measurement 35 years of experiments never made

Since 1989, physicists have trapped a single micro-bubble in water with nothing but sound and watched it do something nobody fully understands: once per acoustic cycle it collapses and emits a flash of light shorter than 200 picoseconds, from gas hotter than the surface of the sun. They call it the star in a jar. The bubble's motion is textbook physics. The light — after 35 years of debate — still is not. Below: a live simulation of the real equation, a 3-D toy that shows the geometry, and at the bottom, the one number nobody ever measured — and our prediction for it, on the record.

1 · THE JAR — one acoustic cycle, live

A live Rayleigh–Plesset integration (the standard equation of bubble dynamics) — R₀ = 4.5 µm, f = 26.5 kHz, the canonical published conditions.
t = 0.0 µsR = 4.50 µm flashes: 0rebound # decay/bounce = state: DRIVEN — watch the settling get interrupted

What to watch. The faint sine is the sound. The bubble swells to ~30 µm, then crashes — ⚡ marks the flash. Then come the afterbounces: the bubble ringing down toward its resting size. Now watch the lower panel — the rebound amplitudes stepping down, bounce by bounce — and watch them get cut off: the next sound cycle seizes the bubble mid-settling, every cycle, forever. In 35 years of published experiments, the settling has never once been allowed to finish. The grey band marks what published data can resolve; the cyan line is our registered prediction (§3) — note it sits below everything ever measured. Press ✂ and the two possible futures draw themselves.

2 · INSIDE THE BUBBLE — a wavefront packet, the turn, and the light pattern it paints

Not one photon — a packet of hundreds of rays (a chunk of wavefront) launched together, each bouncing and turning by Δφ. LEFT: the packet inside the sphere. RIGHT: the boundary unwrapped, rendered as an intensity map — brightness = how much wave energy lands there. This is the light pattern.
|
rays: 160total bounces: 0 wall hits mapped: 0packet energy: 100% state: BOUNCING — the packet paints the pattern

Watch the wave pattern build on the right. Hundreds of rays land, and the map lights up where they concentrate — that is the intensity, the actual pattern of light. Now try the three buttons; same packet, same throw, only the turn angle changes. At 40.000° the wave lands on itself — a handful of blazing spots, a fixed lattice, repeating forever. At 40.020°two hundredths of a degree away — those spots smear into bright arcs before your eyes. At the golden angle (nature's favorite — how a sunflower packs its seeds) the energy never lands twice: a smooth luminous spiral, no lattice, no repeat. One angle decides whether a trapped wave concentrates into fixed hot-spots or spreads into a spiral — and shockingly little of it does. (Without the turn at all, the packet stays trapped in one flat ring: the turn is what fills the sphere.)

3 · THE NUMBER NOBODY MEASURED — a prediction, on the record

Digitize the canonical published curves (we did — sources below) and you can extract two numbers that, remarkably, appear nowhere in the literature as measured values: the bubble's per-bounce decay ratio ρ ≈ 0.74 and its rebound count N ≈ 12 before the next sound cycle interrupts. Run the arithmetic and something jumps out: at that decay rate, the settling needs about 21 bounces to finish — and the drive only ever allows about 12. Every experiment since 1989 has been structurally incapable of watching a bubble finish settling. Not for lack of skill — nobody thought the tail was where the story was.

So here is a falsifiable prediction, registered in advance. Cut the drive at the flash and let the ring-down run free. Textbook physics expects a featureless exponential decay, all the way down — no special amplitude, ever. We predict the settling terminates: the residual oscillation stops at roughly one twelve-hundredth of the initial throw — about 21 nanometers for the canonical bubble (≈40 nm in cold argon, where it settles three times faster). A plateau where the textbook says there is none. The apparatus that decides it is a tabletop: a standard single-bubble cell, a photodiode trigger, and a relay to kill the amplifier — parts, roughly $550. One number, two futures, and either answer is a real result.

Where this thinking comes from. The reasoning behind this prediction — treating loss as reflection, and asking what the settling is really holding — runs through our book Engineering the Impossible (read the opening free). And finding the blind spot no one was testing is exactly what a first-principles engine is for — put your own physics to it at labforge.us. If you run a lab and want to shoot at the number above — message us. First clean result owns the record.
— ProtoDynamics
Sources for every number: Brenner, Hilgenfeldt & Lohse, Rev. Mod. Phys. 74, 425 (2002) — canonical R(t), Fig. 4 · Putterman & Weninger, Annu. Rev. Fluid Mech. 32, 445 (2000) — measured afterbounces, Fig. 5B · Didenko & Suslick, Nature 418, 394 (2002) — energy partition · Gaitan & Crum (1990); Barber & Putterman (1991) — discovery & timing. The simulations integrate the Rayleigh–Plesset equation (van der Waals hard core R₀/8.86); the prediction marked as ours is ours alone — the mainstream model contains no such floor, which is precisely what makes the experiment worth running. © 2026 ProtoDynamics LLC.