The Verification Venue · a thing everyone tells wrong
The Drop That Shatters Itself
A Prince Rupert's drop — a tadpole of glass made by letting a molten bead fall into water — will take a hammer blow on its head and live. Then you barely touch its thin tail and the whole thing flashes to powder. Nothing was added. The violence was frozen inside it the moment it cooled.
The drop is a stress battery. Its outside skin cooled and locked solid first; the inside kept shrinking as it cooled, dragging that skin into fierce compression while the core is left in tension. That stored field is invisible in ordinary light — but it is real, and it is measurable. Toggle the cross-polarized view to see it, then start a crack yourself: on the head, or on the tail.
What just happened
press a button ↑
the head is in compression; the tail is where the tension reaches the skin
Fracture-front speed
— m/s
crack-front / bifurcation velocity
Faster than sound?
—
vs air — but subsonic in glass
The measured band is 1,450–1,900 m/s (Chandrasekar & Chaudhri, 1994, high-speed photography). Above it, the crack keeps branching.
Sets how long the front takes to cross the whole drop. Real drops run a few centimetres of tail.
Click Strike the head and the crack you start dies in the first fraction of a millimetre: it is trying to open a surface that is squeezed shut under ~525 MPa of compression, and a crack cannot run through a material that is being pushed together. Click Snip the tail and everything changes. The tail is a thin filament where the tensile core lies just under a wafer of skin; the tiniest crack there instantly taps the stored strain energy, and a single fracture front accelerates down the whole drop, splitting into more cracks as it goes. The drop eats itself.
"Faster than the speed of sound"
You will read that the crack outruns sound. That is true only against air (≈343 m/s), where the front is Mach 4–5. Inside the glass it is well subsonic: sound (the longitudinal wave) travels at ≈5,400 m/s and the surface Rayleigh wave at ≈3,400 m/s, and a running crack almost never exceeds about 0.6 × the Rayleigh speed (~2,040 m/s) before it must branch instead of speed up — which is exactly the branching you are watching. So the honest sentence is: faster than sound in air, and near the terminal crack velocity for glass — not supersonic in the glass.
Layer 2 — the stress isn't asserted, it's read off the light
Stressed glass is birefringent: it splits polarized light into two rays that travel at slightly different speeds, and the delay between them — the retardation — is proportional to the stress along the path. Between crossed polarizers that delay paints fringes, and counting fringes is reading megapascals. This is the very instrument Aben and colleagues used to get the 700 MPa. The stress-optic law:
Slide between the head skin (~525 MPa) and the tail skin (~700 MPa), the two figures Aben et al. (2016) actually measured.
A longer path through stressed glass stacks up more fringes. The constants below (soda-lime, red light) are fixed and sourced.
—
A mere handful of fringes across a millimetre of glass already means hundreds of megapascals. The stress is not a story about how the drop was made; it is a number you can watch, and it is enormous. Store that much stress as elastic strain energy — u = σ²/2E — and the drop is carrying, per unit volume:
—
That is why the head survives a hammer but the tail detonates: it is not about how hard you hit, it is about whether your crack can reach the tension field. On the head, the thick compressive shell holds the surface shut and suppresses the cone crack a hammer would normally drive. On the tail, the shell is a filament and the crack reaches the stored energy at once — enough energy to manufacture roughly 0.44 m² of new fracture surface per cubic centimetre. It cannot come apart into two pieces. It comes apart into powder.
The check — every number recomputed in front of you
The two headline figures — the surface compression and the crack-front speed — come straight from the peer-reviewed papers (left). Everything the instrument derives from them (right, in green) is recomputed live and by the offline verifier; the green column is the cross-check, not a fresh claim.
| quantity | from the source | computed here |
|---|
Run it yourself: node research/the-drop-that-shatters-itself/verify.mjs. It recomputes every number above from the cited figures and then reads this page's HTML to confirm the drawing and the numbers cannot silently disagree.
What's exactly true, what's idealised, and where the popular story runs ahead of the record
Exactly true (peer-reviewed, primary). The surface layer is in compression of up to ~700 MPa — about 525 MPa on the head, up to ~700 MPa on the tail — measured by integrated (3-D) photoelasticity with a red-LED polariscope: Aben, Anton, Õis, Viswanathan, Chandrasekar & Chaudhri, Applied Physics Letters 109, 231903 (2016). The crack front runs and bifurcates at a critical velocity of 1,450–1,900 m/s: Chandrasekar & Chaudhri, Philosophical Magazine B 70 (1994), high-speed photography. Those two numbers survive checking; everything the page derives from them (mph, Mach vs air, fraction of the Rayleigh/longitudinal speed, traverse time, stored strain-energy density σ²/2E) is elementary arithmetic on them, redone in the verifier.
The load-bearing correction. "Faster than sound" is true only against air (343 m/s). Inside the glass the front is subsonic: longitudinal ≈5,400 m/s, Rayleigh ≈3,400 m/s, and running cracks cap near ~0.6·Rayleigh (~2,040 m/s) before branching. The page never implies the crack is supersonic in glass — it is not; it sits near the terminal crack velocity for glass.
What the record measures — and what it doesn't. The interior tension is real and required by equilibrium — the compression integrated over the thin skin must be balanced by tension integrated over the larger core, so the core tension is lower in magnitude than the skin compression. And it was measured: by the same integrated photoelasticity that gives the 525/700 MPa compression, Aben et al. (2016) report a maximum interior tensile stress of ~300–400 MPa in both the head and the tail. What the record does not support is the widely-copied "664.3 kN head strength" (uncited on Wikipedia); it conflicts with the testing-machine loads in the primary work and is treated as unverified — any head-strength intuition here rests on the compression mechanism, not that number.
Idealised / representative (named free choices). The stress-optic reading uses a relative stress-optic coefficient C ≈ 2.6×10⁻¹² Pa⁻¹ (soda-lime glass is commonly ~2.5–2.7 Brewster) and red light λ = 630 nm; the fringe counts are illustrative of the method (they scale with the path length you choose), not a measured fringe map of a specific drop — but the constants and the conversion are the real ones. The stored-energy estimate uses Young's modulus E ≈ 70 GPa and, for the "new surface" figure, a fracture energy G꜀ = K_Ic²/E ≈ 8 J/m² from K_Ic ≈ 0.75 MPa·m^½. The implied fragment size (tens of microns) is an upper bound on fineness: in reality much of the released energy becomes kinetic energy of flying pieces and heat, so real grains are coarser — the honest point is only that the stored energy is orders of magnitude more than enough to pulverise the glass. The drawn fringe pattern is a schematic of the effect, not a photograph.
The history, corrected. Prince Rupert of the Rhine did not invent these. He brought already-known continental "Dutch tears" (lacrimae batavicae) to England in 1660; Charles II passed them to the Royal Society in 1661, and Hooke described them in Micrographia (1665). The true origin is genuinely murky — the Netherlands around 1656, and possibly Mecklenburg glasshouses before 1625. The popular "Rupert invented the drop" is an attribution error; the record supports "brought to England by," not "invented by."
Two mechanism conflations to refuse. Phone cover glass (Gorilla Glass and kin) is chemically strengthened by ion exchange (larger K⁺ swapped for Na⁺ to crowd the surface into compression) — the same principle of a compressive skin, a different mechanism from the thermal tempering of a Rupert's drop; not "the same process." And it is car side and rear windows that are thermally tempered and dice into cubes on failure; windshields are laminated glass-PVB-glass, a different technology. The same stored-tension release, scaled up, is also behind the spontaneous shattering of tempered-glass façades from nickel-sulfide inclusions. The 1,450–1,900 m/s figure is the crack-front / bifurcation velocity, not the ejection speed of the fragments. This is a bench-safe classroom demo — but eye protection is genuine sense, not theatre.