Why Sand Makes Glass and Salt Never Will
From the first two guides in this rung you already know a glass is a supercooled liquid caught mid-freeze — cooled so fast past its would-be melting point that the atoms never line up into a crystal, locking rigid at the glass transition temperature Tg. But that sharpens a question guide 1 left open: why do some melts freeze so obligingly into a glass while others crystallize almost no matter how hard you quench them? Melt pure silica sand and you get glass with barely any effort; melt table salt or magnesia and you get crystals every single time. Something about the atoms themselves decides.
In 1932 William Zachariasen answered it with one of the most influential short papers in all of materials science. His insight: a glass is built from the same small polyhedra as the corresponding crystal — in silica, the very SiO4 tetrahedron you met in the silicate rung — linked by the same strong bonds, but stacked without the endless repeat of a lattice. Picture a crystal as graph paper and a glass as the same tiles thrown down slightly askew: every tile is still a near-perfect tetrahedron (sharp short-range order survives), yet the Si-O-Si bridges twist to random angles, so there is no long-range order and no repeating unit cell. Zachariasen called this the continuous random network. Remember from guide 1: amorphous never meant chaotic — it means ordered up close, disordered far away.
Zachariasen's Four Rules
So what makes an oxide AmOn a willing glass-former? Zachariasen distilled it into four structural conditions. The gist of his rules is to build a net of oxygen polyhedra that is strongly bonded yet floppy enough to freeze in a jumble rather than snap into a lattice. In plain terms, here they are:
- Each oxygen links to no more than two of the network cations. An oxygen shared by three or four cations would stitch the structure too tightly and force a crystal.
- The network cation sits in a small oxygen polyhedron — a coordination of only three or four oxygens (a triangle or a tetrahedron), never six or eight.
- Those polyhedra share only corners — a single bridging oxygen between neighbours — never edges or faces. (Same corner-only rule as the silicates, and for the same reason: it lets the network flex.)
- At least three corners of every polyhedron are shared, so the units link outward in all directions into a continuous three-dimensional net rather than closing into small isolated rings or chains.
Run the test. Silica passes on every count: an SiO4 tetrahedron is four-coordinate (rule 2), every oxygen bridges exactly two silicons (rule 1), the tetrahedra share corners only (rule 3), and all four corners are shared, comfortably beating the three-corner minimum (rule 4). Now try magnesia, MgO: from the structures rung you know Mg2+ sits in six-fold, octahedral coordination in the rock-salt lattice — rule 2 fails at the first hurdle, its octahedra share edges (rule 3 fails too), and MgO crystallizes the instant it cools. Common salt gets the same verdict. The rules are not arbitrary; each one is a way of keeping the network strong yet loose enough to freeze disordered.
Network Formers: The Atoms That Build the Net
The cations that obey the four rules are the network formers — SiO2 above all, joined by B2O3 (boron oxide) and P2O5 (phosphorus pentoxide), with GeO2 and As2O3 close behind. What do they share? Each is a small, highly charged cation — Si4+, B3+, P5+ — holding its few oxygens in a tight triangle or tetrahedron with strong, directional, half-covalent bonds. Boron makes flat BO3 triangles (three-coordinate); silicon and phosphorus make tetrahedra. These are exactly the units the rules call for, and left to themselves they polymerize into a seamless net. Vitreous (glassy) silica is nothing but SiO4 tetrahedra corner-linked into a random 3-D web — the same brick as quartz, just never allowed to line up.
Physicists capture 'small and highly charged' in one number, the cation field strength — roughly the charge divided by the square of the cation-oxygen distance, z / a^2. It measures how fiercely the cation grips its oxygens. A quick comparison tells the whole story: for Si4+, with a Si-O distance near 1.6 Å, field strength is about 4 / 1.6^2 ≈ 1.6; for a modifier like Na+, with a longer Na-O distance near 2.3 Å, it is only about 1 / 2.3^2 ≈ 0.19 — roughly eight times weaker. High field strength (about 1.3 to 2) marks a former that builds; low field strength (below about 0.4) marks a modifier that loosens. That single ratio sorts the periodic table into the two jobs of a glass.
Network Modifiers: Snipping the Net to Make It Flow
A net of pure silica is magnificent but almost unworkable: with every corner shared it stays as stiff as a solid until roughly 1700 degrees C, far too hot for a bottle factory. The fix is to add a network modifier — an alkali or alkaline-earth oxide such as Na2O, K2O, or CaO. These bring big, weakly-charged cations (Na+, Ca2+) that cannot build a network of their own; instead their oxide donates an extra O2- ion into the melt, and that oxygen chemically snaps a bridge. One Si-O-Si bridge plus one O2- becomes two separate Si-O(-) ends — two non-bridging oxygens, each carrying a negative charge neatly balanced by a nearby Na+ or Ca2+ sitting in a hole in the network.
Every snip depolymerizes the net a little — fewer bridges, more loose ends — and a less connected network flows far more easily. This is the master lever of glassmaking: adding modifier presses the whole viscosity-temperature curve you met in guide 2 downward, dropping the working and softening points by many hundreds of degrees so the glass can be melted, blown, and pressed at practical temperatures. Where pure silica needs about 1700 degrees C, a soda-lime melt is workable near 1000 to 1400 degrees C. (The exact bookkeeping of bridging versus non-bridging oxygens — counting how many bonds each snip breaks — is the whole of guide 4; here just hold the picture that a modifier trades network connectivity for fluidity.)
BREAKING A BRIDGE (one Na2O snaps one Si-O-Si bridge in two)
...Si - O - Si... + Na2O -> ...Si - O(-) (-)O - Si...
^ ^ ^
bridging O two NON-bridging O,
(shared by 2 Si) each balanced by a nearby Na+
per Na2O added: +1 O2- into the melt => -1 bridging, +2 non-bridging
net loosens -> viscosity and working temperature fall
THE THREE ROLES (oxide glasses)
role example oxides cation CN field strength effect on the net
----------- ------------------- --------- --------------- ------------------
former SiO2 B2O3 P2O5 3 or 4 high (~1.3-2.0) builds the network
intermediate Al2O3 TiO2 ZnO PbO 4 to 6 middling (~0.5-1) can join OR loosen
modifier Na2O K2O CaO MgO 6 to 8 low (~0.1-0.4) snaps bridges, fluxesIntermediates, and Reading a Real Glass
Between the two clean jobs sits a fuzzy middle class, the intermediates — Al2O3, TiO2, ZnO, PbO. Their cations have middling field strength and can play either part depending on the company they keep. Alumina is the classic two-faced example: drop Al2O3 into a soda-silica melt and each Al3+ will take four oxygens and slot into the network as an AlO4 tetrahedron, behaving as a former — but because AlO4 carries a spare negative charge, it pulls a Na+ over to balance it, quietly consuming a modifier and healing non-bridging oxygens back into bridges. That is why a touch of alumina stiffens a glass and makes it far more durable, and why 'former versus modifier' is a role an atom plays, not a fixed label it wears.
Now you can read a real glass recipe like a sentence. Ordinary soda-lime-silica glass — windows, bottles, the cheapest and most common glass on Earth — is roughly 72 wt% SiO2 (the former), 14 wt% Na2O (the modifier that makes it workable), and 10 wt% CaO. Why the lime? Soda alone would leave the glass water-soluble; the calcium re-stiffens the net and restores chemical durability — CaO is the 'stabilizer' that rescues what the soda gave away. Borosilicate glass (Pyrex, Kimax) tells a different story: about 80 wt% SiO2 and 13 wt% B2O3, with only a little soda. With boron sharing the network-forming job and few modifiers to loosen it, borosilicate has a very low thermal-expansion coefficient — near 3.3 x 10^-6 per degree C against about 9 for soda-lime — so it barely swells when heated and shrugs off the thermal shock that shatters an ordinary tumbler filled with boiling water.
So Zachariasen's rules give you a lens for every glass you will meet: find the formers building the random net, find the modifiers snipping it loose to control viscosity, and watch the intermediates hedge in between. The very same net also explains the failures ahead — let a glass linger too long in the crystallizing danger zone and the net finds its lattice after all (devitrification), while doing that crystallization on purpose, with a seeded nucleating agent, yields the tough, low-expansion glass-ceramics of guide 5. But first, guide 4 zooms all the way in on the single bond this guide kept pointing at: the bridging oxygen, and what its snipping really costs.