Tempered Glass - 3D Tempering Line

Quench a hot glass plate, see the stress freeze in with polarised light, then break it and count the dice.

Switch to 2D

Presets

Playback

Plate

Thickness, furnace temperature and the temperature at which the glass stops flowing.

Quench and optics

Cooling intensity decides the stress; path length only changes the fringe count.

Camera

Layers

Quality

Model checks

The science

How a quench locks stress into glass

Hot glass above its freezing (strain-point) temperature Tf flows and carries no stress. Air jets cool the faces first: the skin passes Tf and becomes elastic while the core is still soft. The core then cools and tries to shrink by alpha (Tf - Tair) but is held by the rigid skin, so the core ends in tension and the skin in compression. The through-thickness profile is close to a parabola, sigma(z) = sigma_c (1 - 3 (z/H)^2), whose force balance integral sigma dz = 0 forces the surface value to be about -2 sigma_c. Typical tempered glass: 90-150 MPa compression at the surface, 40-70 MPa tension in the core.

The model in this lab

The temperature field comes from the heat equation in a slab, rho c dT/dt = k d2T/dz2, with a surface loss -k dT/dz = (hc + h_rad(Ts)) (Ts - Tair) (convection plus linearised radiation, emissivity 0.84), solved by finite differences. Layers colder than Tf are elastic: each time step the frozen layers receive thermal-strain increments alpha dT, the common in-plane strain increment follows from sum(w d sigma) = 0, and d sigma = E' (d e - alpha dT) with E' = E/(1 - nu) (E = 70 GPa, nu = 0.22, alpha = 9e-6 /K). More cooling (larger Biot number Bi = hc H / k) means a larger temperature difference at the moment the core freezes, so more stress; thicker glass at the same jets also gains stress.

Polariscope

Photoelastic glass is birefringent in proportion to the stress: the path difference between the two polarisations is delta = C sigma L with the stress-optic coefficient C = 2.7e-12 /Pa and light path L. Between crossed polarisers at 45 degrees the intensity of wavelength lambda is sin^2(pi delta / lambda), so each colour has its own fringes. You look along the plate edge (the in-plane stress acts across the light path); the neutral lines, where stress changes sign, are dark.

Why it shatters into dice

Breaking the skin releases the stored elastic energy per unit area U = integral sigma^2 / (2 E') dz. A running crack keeps branching while that energy pays for new surface; with an effective crack energy Gamma (8 J/m2 here, including roughness) and half of U available, square dice of side d = 2 Gamma h / (0.5 U) result. Typical output for 6 mm glass: dice of 3-5 mm, and the standard EN 12150 asks for at least 40 fragments in any 50 x 50 mm window (d at most 7.9 mm). Annealed glass has almost no stored energy, so it breaks into a few large shards. The Prince Rupert drop is the extreme case: quenched in water, its surface compression is several hundred MPa; the head withstands a hammer, yet snapping the thin tail sends a fracture front through the whole drop at about 1.5 km/s, turning it to powder.

What is simplified

Instant freezing at one temperature replaces the full viscoelastic structural relaxation of glass (real tempered stress is somewhat different); constant glass properties; one-dimensional symmetric cooling, so edge and corner effects, quench marks and the roller-wave distortion are ignored; the drop is a slab of half-thickness equal to its head radius (a sphere has a different surface-to-core ratio, which is why the printed ratio differs); fragment size comes from an energy balance, not from crack-branching dynamics; the crack pattern is a Voronoi picture at the computed scale. Plate sizes and the drop are drawn for clarity; slow-motion factors are noted on screen.

What am I looking at
  • Furnace and rollers: the plate leaves the furnace glowing red; colour follows its surface temperature.
  • Quench box: upper and lower plenums with a nozzle array; white streaks are the air jets.
  • Punch table: a spring punch on one corner; the crack pattern spreads from the impact point.
  • Polariscope bench and screen: the edge view between crossed polarisers; the green line is the fringe order.
  • Graphs: temperature history, stress through the thickness (now, final, ideal parabola), fringes, and fragment size against cooling intensity.
What to try
  • Press Quench with "Tempered 6 mm" and watch the stress column: first the skin goes into compression, then the core swings into tension as it cools. Open the Polariscope column to see the colour fringes grow.
  • Then press "Punch / snap test": about 180 dice in the 50 x 50 mm window. Pick "Annealed (slow cool)" and punch again: a handful of large shards.
  • Drag "Cooling intensity" up and down and watch the fragment-size graph cross the EN 12150 line.
  • Compare 3 mm with strong jets against 12 mm: thicker plates reach higher stress at the same jets.
  • Choose the Prince Rupert drop and press Snap: the fracture front runs from tail to head.
Loading model...
Loading Blender model...
Temperature at the centre and the surface
Stress through the thickness (now, final, ideal parabola)
Polariscope: colour fringes and fringe order
Fragment size vs cooling intensity (EN 12150 limit)
Made with 🟠 Blender · Three.js