Glass transition

Why does window glass soften gradually over a range of temperatures instead of melting sharply at one, as its crystalline twin does?

Ice has a melting point. Put a block of it on a rising thermometer and at one temperature it turns to water, the thermometer refusing to climb until the last of it is gone.

Window glass has no such moment. Heat it and it grows gradually less rigid across hundreds of degrees — stiff, then workable, then runny — with no temperature at which the solid ended. Nor is this a difference of substance: silica has crystalline forms that melt as sharply as ice. The same oxide, arranged differently, has a melting point; arranged as glass, it does not. So what does a crystal have that lets it name a temperature, and what has the glass lost?

Ask what happens at a melting point. A crystal is a lattice: every atom in a repeating position, the whole thing one ordered structure. Melting destroys that order, and because the order is a single connected arrangement it cannot be dismantled a little at a time. Below the melting temperature the crystal is the lower-energy state, above it the liquid is, and at the melting point the two are exactly balanced and coexist — which is why the temperature stalls while the heat you add goes into breaking the arrangement. That stall is the latent heat, the signature of a genuine transition between two states.

Now ask the same of glass, and the argument fails at the first step. There is no lattice. A glass has short-range order — each silicon still holds its oxygens — but no long-range arrangement to destroy. Nothing is waiting to break, so there is nothing for a latent heat to pay for.

Then what is changing as the glass warms? Only how readily the network can rearrange. In a hot silicate melt, bonds break and re-form constantly and the material flows. Cool it and that rearranging slows — and here is the point that decides everything — extraordinarily steeply, by many orders of magnitude across a modest range. Nothing discontinuous happens. It simply gets more sluggish, without limit.

If that is all, why does the glass ever stop being a liquid? Put two clocks side by side: how long the network takes to rearrange appropriately to its current temperature, and how fast you are cooling it. While rearrangement is the faster, the material keeps up and is a liquid at equilibrium, merely a very viscous one. As cooling continues the first clock lengthens catastrophically while the second does not. At some point they cross, and the material can no longer finish rearranging before the temperature has moved on. Its arrangement is left behind — frozen in, appropriate to a slightly warmer material — and what you have is a solid.

That crossing is the glass transition, and notice what kind of event it is. Not a competition between two states over which is more stable, but a race between a rate and a rate. That places it on the kinetics side of the divide, where the melting point sits squarely on the thermodynamics side, and almost every confusion about glass comes from filing it under the wrong one. The physical ageing of plastics, examined elsewhere in this collection, takes this crossing as given and asks what a glass does in the years after it; here we ask why the crossing is not a melting point at all.

Once you see it as a race, the softening range explains itself. A melting point is a property of the substance; a crossing point depends on both clocks, and one is yours. Cool the same melt ten times faster and the crossing happens slightly higher, because rearrangement runs out of time sooner. There is no single temperature to find, because "is this a solid?" has no answer independent of how quickly you ask. The glassblower's working range is not vagueness in the material — it is what a continuous viscosity curve looks like when you insist on labelling parts of it, which is why the trade defines softening and annealing points by viscosity rather than by any transition.

That gives a decisive test whose virtue is that it uses the crystal as its own control. If the glass transition were a true phase transition, its temperature would be a material constant and it would absorb a latent heat. If it is a kinetic crossing, it must shift with the rate you run the experiment at, and show no latent heat. Run a calorimeter over the same glass at rising heating rates: the transition moves upward, modestly but reliably, and what appears is a step in heat capacity, not a peak of absorbed heat. Run it on the crystalline form and the melting point does not move while the latent-heat peak is unmistakable. What would refute the account? A transition temperature that stayed put across a hundredfold change in rate, or a latent heat at it.

One correction: the classroom claim that glass is a liquid, evidenced by old panes thicker at the bottom, is wrong — below its transition a glass is a solid on every practical timescale, and the uneven panes are artefacts of manufacture. What is true is weaker and more interesting: a glass is out of equilibrium and keeps packing very slowly, which is not flowing.

Two honest qualifications. First, the transition being kinetic does not make it arbitrary. It moves only weakly with rate — a few degrees per tenfold change in cooling or heating rate is the magnitude usually quoted — so a stated transition temperature is serviceable as long as the method is stated with it.

Second, whether anything thermodynamic hides underneath is genuinely unsettled. Extrapolating the supercooled liquid's entropy below the observed transition suggests it would fall beneath the crystal's at a finite temperature, which cannot be right, and whether that points to a true underlying transition to an ideal glass or merely to a failed extrapolation is still argued. The kinetic account describes what is measured; it does not close that question.

Every term the collection defines is gathered in the glossary.