Tempered glass
A Socratic walk-through of tempered glass — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does glass locked in permanent internal stress break less easily than unstressed glass?
Stress is what breaks things. So a pane deliberately manufactured with enormous permanent stress frozen into it — surface squeezed, interior pulled — ought to be nearer failure than a pane left alone. Tempered glass is that pane, and it takes several times the load before it breaks.
Something in the intuition must be wrong. Either stress is not simply bad, or glass does not break the way we assume. It turns out to be the second.
Reasoning it through
REASONING #Ask first what actually limits the strength of a sheet of glass. The chemical bonds in silica are strong; a flawless glass fibre can carry loads in the gigapascal range. Ordinary window glass fails at a small fraction of that, and worse, at unpredictable loads — two identical panes can differ severalfold. A material whose strength is a property of its chemistry does not behave like that.
What varies between the two panes is not chemistry but damage. Every glass surface carries microscopic flaws: scratches from handling, contact points, nicks invisible to the eye. Griffith's insight was that a crack concentrates stress at its tip in proportion to how sharp and deep it is, so a flaw is a lever multiplying the applied load locally until the bonds at the tip give way. Glass at room temperature has no way to blunt that tip — no plastic flow, no dislocation motion — so once the tip starts moving, nothing stops it.
Two consequences follow. Glass strength is a property of the worst flaw, not of the material. And glass breaks in tension: a crack only propagates if something pulls its faces apart. Squeeze a flaw and it stays shut.
That reframes the problem. To strengthen glass you do not need better glass; you need to stop the flaws already there from opening. And there is a way to do that without touching them: put the surface into permanent compression, so any applied tension must first spend itself cancelling that compression before the crack tip feels any pull at all.
How would you build such a state? Heat the pane until it is soft enough to flow, then blast both faces with air. The surfaces cool and become rigid first, while the interior is still hot and fluid, so they set at a large size. Then the interior cools in its turn and tries to contract — but it is attached to rigid faces that will not shrink with it. The core ends up pulled outward, in tension, and pulls the faces inward, into compression. The stress is locked in because the geometry that would relieve it was frozen before the contraction happened.
The two must balance: with nothing pushing from outside, the pane's internal forces sum to zero, so compression near both faces is paid for by tension in the middle. For the roughly parabolic profile quenching produces, working the integral through gives central tension at about half the surface compression — the interior is genuinely in a dangerous state, just not one exposed to any flaw.
So the payoff and the price in one sentence: the pane is stronger because a load must overcome the surface compression before it can open anything, and more dangerous when it does fail because the core's stored elastic energy is released the instant a crack reaches it.
That price explains the peculiarities. Tempered glass does not crack; it dices, the crack branching repeatedly as it runs through the energy-rich core into thousands of small blunt fragments — which is why it is the safety glazing in car side windows. It cannot be cut or drilled afterwards, because any cut reaches the tension core. And it occasionally shatters spontaneously years later, most often traced to nickel sulfide inclusions that slowly change phase and expand inside the tension zone until they start a crack from within.
The analogy
THE ANALOGY #Think of a barrel bound with iron hoops driven on hot. As they cool they contract and squeeze the staves together permanently. The staves are not stronger wood, and the joints between them are as imperfect as before — but they cannot open, because the hoop holds them shut, and any outward push from the contents must work against that grip before a gap can appear.
the hoop is a separate part you could cut off, leaving sound staves behind, whereas in tempered glass the squeezing skin and squeezed core are one continuous body — you cannot remove the compression without releasing the tension that pays for it, which is why a cut destroys the pane.
Clarifying the model
THE MODEL #The most important correction is that tempering does not make the material stronger. The glass is chemically identical before and after, and the flaws are still there, the same size. What changes is how much external load must be applied before those flaws experience net tension. A tempered pane and an annealed pane fail at the same local stress at the crack tip; they simply need different applied loads to reach it.
The practical corollary is that the protection lives in a thin surface layer. Damage penetrating past it — a deep scratch, a drilled hole, a chipped edge — does not merely weaken the pane in proportion; it can reach the tension core and start a crack the whole stored energy of the pane is waiting to drive. Which is why edge damage matters far more in tempered glass than in ordinary glass.
Two qualifications. Thermal quenching is not the only route: chemical strengthening swaps surface sodium ions for larger potassium ones from a molten salt bath, crowding the surface into compression with no thermal history. That gives higher compression in a much thinner layer and breaks into large shards rather than dicing — same principle, different depth profile, different failure mode. And the numbers are specification thresholds rather than properties of glass: the usual industry bar for "fully tempered" is a surface compression of roughly 70 MPa, heat-strengthened glass sits well below that, and the strength gain over annealed glass is conventionally quoted as around fourfold.
A picture of it
THE PICTURE #How to readRead left to right as a journey straight through the pane, face to face. The curve dips deep below zero at both faces — that is the compression holding every surface flaw shut — and rises above zero through the middle, where the glass is being pulled apart. The areas above and below zero must balance, because nothing is loading the pane from outside; that balance is why the central tension comes out at about half the surface compression, and why you cannot have one without the other. Applied tension shifts the whole curve upward: the pane survives until the face crosses back above zero, and fails catastrophically if anything reaches the hump in the middle.
What became clearer
WHAT CLEARED #Glass does not break because the material is weak; it breaks because it is full of flaws that cannot be blunted and a crack tip that cannot be stopped. Once that is clear, permanent stress stops being a liability and becomes a tool: putting the surface in compression means every applied load is first spent undoing it, and the flaws never feel a pull. The core's tension is not a side effect to be tolerated — it is the exact price of that compression, in equal and opposite measure, and it is what makes the eventual failure so sudden.
Where to go next
ONWARD #- Why glass fails at lower loads under sustained rather than momentary stress, and what water at the crack tip has to do with it.
- Why the depth of the compressive layer, not its peak stress, decides how much damage a phone screen survives.
Key terms
TERMS #| Term | What it means |
|---|---|
| Residual stress | stress remaining in a body with no external load applied, necessarily summing to zero across the section. |
| Griffith flaw | a microscopic surface crack that concentrates applied stress at its tip and so sets the real strength of a brittle material. |
| Annealed glass | glass cooled slowly enough that no significant residual stress is retained; the untreated baseline. |
Every term the collection defines is gathered in the glossary.