A Solid That Forgot to Crystallize
You already know the SiO4 tetrahedron, and you have seen how, in quartz, those tetrahedra corner-share into a framework that repeats perfectly in every direction — a crystal. A glass is built from the very same tetrahedra, joined at the very same corners, but the framework never repeats. Picture sliding a window pane's atoms around in your mind: each silicon still sits at the centre of four oxygens, yet there is no long-range pattern, no lattice you could tile out to infinity. That aperiodic, frozen-liquid arrangement is the glassy (amorphous) state.
And a glass is still a ceramic. 'Ceramic' names a family bonded by strong ionic-covalent bonds — inorganic and non-metallic — not a shape or a recipe. But notice that the route is different: most ceramics you have met are shaped from powder and firmed up by firing without ever melting (remember: sintering needs no melting), whereas a glass is melted fully to a liquid and then cooled. Same family, opposite journey.
Freezing a Liquid Without Crystallizing
Picture cooling a pot of molten silica. Crystallizing means every tetrahedron must find its exact lattice site — atoms have to migrate, rotate, and line up. That takes two things: nucleation (a first tiny crystal seed appears) and growth (the seed spreads). Both need atoms to move, and moving gets harder as the melt thickens. Glass formation is really just winning a race: cool fast enough that the melt turns stiff before crystals can nucleate and grow, and the atoms are caught out of position — frozen mid-shuffle. The still-liquid-structured melt on the way down, already below its freezing point, is a supercooled liquid.
volume
^
| o liquid (hot)
| _,-'
| supercooled _,-'
| liquid _,-'
| _,-'
| GLASS ___________ _,-'
| line (elbow at Tg) ...at Tm the melt can
| instead crystallize:
| drop off a cliff |
| CRYSTAL v
| line ____________________________________ ___
| (dense, ordered; below the glass line)
+----------+-----------------------------+----------> T
Tg TmSome liquids are easy to trap and some are nearly impossible. Silica is a champion glass-former because its melt is already fantastically viscous — the tetrahedral network barely flows — so crystals cannot assemble in time even under slow cooling. A pure metal is the opposite: its atoms are small and slippery, so unless you quench it at millions of degrees per second it crystallizes every time. The lesson is that high melt viscosity is what makes glass formation easy, and we will trace that viscosity curve closely in the next guide.
Tg: Where the Liquid Freezes In
As the supercooled melt cools, its viscosity climbs astronomically — not by a little but by many powers of ten. When it reaches roughly 10^12 Pa s (about a trillion times stiffer than honey), the atoms can no longer rearrange within any reasonable time, and the material behaves as a rigid solid. The temperature where this hand-off happens is the glass transition temperature, Tg. Above Tg you have a very thick liquid; below Tg you have a glass.
Here is the crucial honest point: Tg is not a fixed melting point. A crystal melts at one sharp temperature Tm, set by thermodynamics. Tg, by contrast, is set by kinetics — by how fast you cooled. Cool faster and the melt runs out of time sooner, freezing in at a slightly higher Tg; cool slower and Tg drops. Move the cooling rate and you move Tg by tens of degrees. A glass is therefore a non-equilibrium, metastable solid: it is not a phase on any equilibrium phase diagram, only a liquid that got stuck.
The Random Network: Formers and Modifiers
In 1932 W. H. Zachariasen gave us the picture we still use: a glass is a continuous random network of the same coordination polyhedra found in the crystal, linked at their corners into a three-dimensional web that fills space without ever repeating. Zachariasen's rules tell you which oxides can do this — the ones whose small, highly-charged cations sit in small triangles or tetrahedra of oxygen and share only corners, never edges or faces. We will unpack those rules in guide 3.
The atoms play distinct roles. Network formers — SiO2, B2O3, P2O5 — are the ones that build the network; each links to its neighbours through shared bridging oxygens, oxygens that belong to two polyhedra and stitch the web together. Pure silica is all bridges, which is exactly why it is so viscous and so hard to melt. Network modifiers — Na2O, CaO, K2O — do the opposite: dropped into the melt they snap a bridge, turning one shared bridging oxygen into two dangling non-bridging oxygens, each capped by a sodium or calcium ion. Every snapped bridge is a cut strand in the net.
Why bother cutting the net? Because pure silica is impossibly stiff to work — you would have to shape it above 2000 degrees C. Add about 15 mol% Na2O, snap enough bridges, and the working temperature falls toward 1000 degrees C, cheap enough to make bottles by the billion. That is the whole trade of practical glassmaking: modifiers lower the melting and working temperatures (and the viscosity), but because they replace strong bridges with weaker non-bridging bonds, they also cut chemical durability and raise thermal expansion. Bridging versus non-bridging oxygen is the lever behind almost every property, and guide 4 is devoted to it.
Two Glasses You Use Every Day
The commonest glass on Earth is soda-lime-silica — roughly 72 wt% SiO2, 14 wt% Na2O, 10 wt% CaO — the stuff of windows, jars, and bottles. The soda (Na2O) is the modifier that makes it cheap to melt; the lime (CaO) is added back to restore the chemical durability that a soda-only glass would lack (soda-only glass is 'water glass' and literally dissolves in water). Its weakness is a high thermal expansion coefficient, near 9 x 10^-6 per degree C: heat one face faster than the other and the mismatched expansion cracks it. Pour boiling water into a cold, thick tumbler and this is exactly what fails.
Borosilicate glass — the Pyrex family — swaps much of that soda for B2O3, a second network former. B2O3 keeps the network well-bridged, so the expansion coefficient falls to about 3.3 x 10^-6 per degree C, roughly a third of soda-lime's. That is why a borosilicate beaker or coffee pot shrugs off a thermal shock that would shatter a window: with so little expansion, uneven heating builds far less stress. Push further to pure fused silica and the expansion drops near 0.5 x 10^-6 per degree C — superb, but its sky-high melting point makes it expensive.
Stress, Annealing, and Glass That Crystallizes on Purpose
When a hot glass object cools, its skin freezes before its core; the two end up locked in disagreement, leaving frozen-in residual stress that can crack the piece days later. The cure is annealing: hold the glass near its annealing point (viscosity about 10^12 Pa s) long enough for the stresses to relax, then cool slowly. You can also weaponise the same physics: thermal tempering deliberately chills the surface of hot glass with air jets so the skin sets first, and when the core finally shrinks it squeezes the skin into permanent compression. Since a crack must be pulled open in tension to grow, that compressive skin is a shield — it is why tempered glass is several times stronger and crumbles into blunt pebbles instead of daggers.
- Melt the network formers together with modifiers until the batch reaches a workable viscosity — the modifiers snap bridges to bring the melting temperature within reach.
- Homogenise and 'fine' the melt, letting trapped bubbles rise out so the glass is clear and uniform.
- Shape it while it is pyroplastic — blow, float on molten tin, press, or draw — working within the viscosity window where it holds shape but still flows.
- Cool through Tg fast enough to dodge crystallization, so the melt freezes into a glass rather than growing crystals.
- Anneal near the annealing point to relax residual stress — or, for strength, temper the surface into compression instead.
Crystallization is not always the enemy defeated — sometimes it sneaks back, and sometimes you invite it. If a glass is held too long in the temperature band where crystals nucleate and grow, unwanted crystals appear, clouding and weakening it; this failure is devitrification ('un-glassing'). But tame the same process and it becomes a triumph: a glass-ceramic is a glass melted and formed the easy way, then deliberately reheated with a nucleating agent seeded throughout so that countless tiny crystals grow in unison. The result is a dense, fine-grained, largely crystalline body — the near-zero-expansion cooktops and cookware that take a burner flame straight onto ice. Same crystallization; the difference between ruin and engineering is control.