A liquid caught mid-freeze
The last guide built silica up from a single Lego brick: the SiO4 tetrahedron, one silicon atom cradled by four oxygens, each oxygen shared with a neighbour so the whole thing knits into a three-dimensional net. If those tetrahedra line up in a perfectly repeating pattern, you get crystalline quartz — a crystalline solid with the long-range order of a tiled floor stretching to the horizon. Glass is made of the very same tetrahedra, connected in the very same corner-sharing way, with one difference that changes everything: the net is tangled at random. There is no repeating lattice, no long-range order at all. Glass is an amorphous solid.
So the glass structure keeps the short-range order — every silicon still sits at the centre of its four oxygens, every bond still the right length — while losing all long-range order. Picture a crowd. In a crystal, everyone stands on marked spots in tidy rows, and once you know one person's spot you can predict everyone's for a mile. In a glass, everyone still holds hands with exactly the right neighbours, but the crowd froze mid-shuffle: locally sensible, globally chaotic. That frozen-crowd picture is the whole idea, because it tells you what glass really is — a liquid whose jumbled arrangement got locked in place before it could ever tidy itself into a crystal.
The glass transition: freezing without crystallising
Here is how a liquid decides to become glass. Cool any melt and its atoms slow down and pack closer, so its volume shrinks steadily. At the melting point Tm, one of two things can happen. If the atoms have time and freedom to organise, they snap into a lattice all at once — that is crystallising, and it shows up as a sudden step down in volume, plus a burst of released latent heat. But if you cool too fast, or the melt is a tangled silica net that simply cannot reshuffle quickly enough, the liquid sails right past Tm without crystallising. It is now a supercooled liquid, still liquid-structured but below where it 'should' have frozen.
Keep cooling and the supercooled liquid gets thicker and thicker — its viscosity climbs astronomically. At some point the atoms are moving so sluggishly that they can no longer rearrange in the time you are giving them: the structure simply stops evolving and locks in place. That temperature is the glass-transition temperature, Tg. It shows up not as a sharp step but as a gentle bend in the volume-versus-temperature curve, where the melt quietly stops behaving like a liquid and starts behaving like a solid — with no phase change, no latent heat, no crystallising. The disordered liquid structure is now frozen in. That bend is the birth of glass.
COOLING A MELT -- two destinies (volume shrinks as it cools)
high V | liquid \
| \
| \ (supercooled liquid)
| \____
| Tg \_____ GLASS <- smooth BEND: the
| \____ disordered liquid
| crystal ____________ structure, frozen
| ________| <- Tm
low V |____| CRYSTAL ^ sharp STEP: atoms snap into a lattice
+-------------------------------------> temperature
Tg Tm
Cool slowly, and it can organise -> CRYSTAL (step at Tm)
Cool fast, or too tangled to organise -> GLASS (bend at Tg)No melting point, just a viscosity ramp
The deepest practical fact about glass falls straight out of that bend: glass has no single melting point. A metal is icy-solid one degree and a runny puddle the next — a sharp on/off switch. Glass instead softens over a wide temperature range as its viscosity slides smoothly through many powers of ten. Water flows at about 10^-3 Pa-s and honey at roughly 10 Pa-s; molten glass ready to gather is around 10 Pa-s, thickens to about 10^3 Pa-s where a glassblower can work it, stiffens to roughly 10^12 Pa-s near Tg where it will hold a shape, and at room temperature it is past 10^18 Pa-s — so rigid the word 'liquid' has lost all meaning. This gentle ramp is exactly why glass can be blown, drawn, and pressed: there is a long, forgiving window of 'soft but not runny' that a crystalline solid never offers.
This viscosity ramp is also cousin to something you met on the polymer rungs. A thermoplastic near its own glass transition shows the same in-between behaviour — part springy solid, part slow-flowing liquid — the time-and-temperature-dependent response called viscoelasticity. Glass and a warm plastic are both amorphous, so both go through a glass transition rather than a crisp melt. The difference is only the temperature: for window glass Tg sits well above 500 degrees C, while for a rubber band it can be below room temperature, which is exactly why the rubber is floppy and the window is not.
Formers and modifiers: tuning the melt
Pure silica glass is magnificent — chemically tough, optically clear, barely expanding when heated — but it is a beast to work, needing furnace temperatures near 1700 degrees C because that continuous SiO4 net is so hard to loosen. Silica is the archetypal network former: an oxide (like B2O3 too) whose bonds actually build the glassy net. The fix is to throw in a network modifier — usually soda (Na2O) from ordinary salt-cake or soda ash. Sodium ions do not join the net; they crash into it, snapping some Si-O-Si bridges and leaving behind dangling 'non-bridging' oxygens. Each broken bridge is one less strand holding the net together, so the whole thing loosens, and both the melting point and the viscosity plummet.
That is how you get ordinary soda-lime glass — roughly 72 percent silica, 14 percent soda, 10 percent lime — the cheap workhorse of windows and bottles, meltable and formable near 1000 degrees C instead of 1700. But soda alone would make the glass slowly dissolve in water (that is literally what 'water glass' is), so lime (CaO) is added as a stabiliser to restore durability. This former-plus-modifier recipe is a design dial: change the oxides and you retune the glass. Swap in boron and you get borosilicate (Pyrex), whose low thermal-expansion coefficient near 3 x 10^-6 per degree C — about a third of soda-lime's — lets it survive being yanked from oven to countertop without cracking.
Strong yet shattering — and how tempering cheats
By its bonds, glass should be tremendously strong — its stiffness, a Young's modulus near 70 GPa, is right up with aluminium, and a pristine glass fibre really can pull to gigapascals. Yet the pane in your window fails at a feeble 40 to 50 MPa, a hundredth of what the bonds allow. The culprit is not the bonds but the tiny scratches and flaws on the surface. Being amorphous, glass has no dislocations and no slip to blunt a sharp crack — nothing gives, so the tip of a flaw magnifies the stress enormously (stress concentration) until a crack runs at the speed of sound and the whole thing lets go. That is textbook brittle fracture, and it is why glass is strong in compression, where cracks are squeezed shut, but weak in tension, where they are pulled open.
So the strength you measure is really the strength of the worst flaw, which is why a single 'strength of glass' number is a lie and glass makers quote a flexural strength (the modulus of rupture from a bend test) surrounded by huge scatter — two identical-looking panes can differ twofold. Taming that scatter statistically is the job of Weibull statistics, the whole subject of the next guide, so hold the thought. The clever engineering answer, though, is not to hunt down every flaw but to make the surface refuse to pull the flaws open in the first place.
That is tempered glass, and the move is the same quench-then-relax logic you saw for steel, turned to a different end. Heat the pane until it softens, then blast cold air on both faces. The surfaces freeze rigid first; a moment later the still-hot interior cools and tries to contract — but the solid skin will not let it, so the interior is left pulling the skin inward. The result is a pane whose surface is locked in permanent compression and whose core is in balancing tension. Now a surface crack cannot even begin to open until the applied load has first cancelled all that built-in compression, so tempered glass runs three to five times stronger. And when it finally does fail, the stored energy dices it into a hail of small blunt cubes instead of long guillotine shards — which is exactly why it lines your car's side windows and shower door.
- Heat the finished pane above its softening range (around 600 degrees C) so the whole thing is soft and stress-free.
- Blast both surfaces with jets of cold air. The skin cools and freezes solid while the interior is still hot and soft.
- As the trapped interior finally cools, it contracts and pulls the already-rigid skin inward — putting the surface into permanent compression and the core into tension.
- A crack must now overcome that surface compression before it can open, so the pane is far stronger; if it ever fails, it crumbles into small, safe cubes.
Where glass sits in the ceramic family
Glass is only one branch of the ceramic world — the amorphous one. Its crystalline cousins, the engineering ceramics like alumina, zirconia, and silicon carbide, come in guide 5, along with the humble porcelain that is really glass and crystal mixed. You can even blur the line on purpose: hold a glass just above Tg and coax it to nucleate and grow a fine crop of tiny crystals throughout, and you get a glass-ceramic — the tough, near-zero-expansion material in stovetop cooktops and old Corningware, born a glass and finished a ceramic. And the purest triumph of glass is invisibility: an ultra-clean silica optical fibre so free of flaws and impurities that light runs down it for kilometres, carrying the internet under the oceans.
Step back and the picture is simple. Glass is a liquid whose disordered structure was frozen in through the glass transition rather than tidied into a crystal, which gives it a viscosity ramp instead of a melting point and makes it beautifully formable. It shares the whole ceramic family's fatal habit — strong but brittle, strong in compression and weak in tension, failing from its worst surface flaw. Tempering does not fix that habit; it disguises it, holding the flaws shut with built-in compression. The next guide meets the brittleness head-on and shows why, when failure hides in the worst flaw you cannot see, you stop quoting one strength and start quoting the odds.