From a list of distances to an actual picture
Guide 3 handed you a powerful but oddly flat piece of information. The radial distribution function of silica glass has a sharp first peak at about 1.6 angstrom (that is the silicon-to-oxygen bond), a second peak near 2.6 angstrom (oxygen-to-oxygen across a tetrahedron), and then it washes out into featureless ripples — proof of crisp short-range order with no long-range periodicity, exactly the fingerprint of the amorphous state. But an RDF is a statistical average over every atom at once; it tells you how far apart neighbours sit without ever showing you the arrangement they sit in. It is a shadow, not the object.
In 1932 William Zachariasen asked the question that turns the shadow back into a solid. A silica glass is mechanically rigid like crystalline quartz, made of the same atoms in the same proportions, with the same tight nearest-neighbour bonds — yet it never repeats. What arrangement can be both? His answer is the continuous random network (CRN): take the exact building block of the crystal, the SiO4 tetrahedron, and link the tetrahedra corner to corner into a single connected framework that fills all of space — but let the angle at each shared corner vary a little at random, so the network never settles into a repeating pattern. Think of crystalline quartz as wallpaper, one motif stamped at identical spacing forever; the CRN is the very same tiles, but the mortar joints between them flex by a few degrees each, so the wall covers the room without the pattern ever coming back.
The network up close, and what makes a good former
Zoom into vitreous silica and you can see exactly where the order lives and where the disorder hides. Inside each tetrahedron everything is stiff: the Si-O bond length is pinned near 1.62 angstrom, the O-Si-O angle sits at the tetrahedral 109.5 degrees, and the silicon coordination number is a rigid 4 — this is the strong, directional covalent bonding doing its job, and it is why the first RDF peak is razor-sharp. The freedom is entirely in the linkage BETWEEN tetrahedra. Each corner oxygen is a bridging oxygen, shared by two silicons, and the Si-O-Si angle across that bridge is not fixed at all: it spreads over a broad range from roughly 120 to 180 degrees, peaking near 144 degrees. That one soft, floppy angle, sampled independently at every bridge, is the entire source of a glass's randomness — a hinge at every joint of an otherwise rigid framework.
The oxides that build such networks — SiO2, B2O3, GeO2, P2O5 — are the network formers, and Zachariasen distilled why they work into a short set of geometric rules. The recurring theme is small, corner-sharing polyhedra: a low cation coordination (3 or 4) keeps the units open and floppy enough to hinge, and sharing only corners (never edges or faces) lets the joints swing freely without forcing the whole array into registry. Formers pass the rules; oxides like Na2O or CaO fail every one of them and cannot build a network on their own.
- Each oxygen is linked to no more than two of the network cations — so oxygens act as two-armed bridges, not crowded hubs.
- The number of oxygens around each cation is small — 3 (a triangle, as in B2O3) or 4 (a tetrahedron, as in SiO2).
- The oxygen polyhedra share only corners with one another, never edges and never faces.
- At least three corners of each polyhedron are shared, so the units link into a continuous three-dimensional network rather than isolated clusters.
Cutting the network: modifiers and non-bridging oxygen
Pure fused silica is a superb glass, but a punishing one to make: with every oxygen tying two silicons together, the network is fully polymerized, and it stays stubbornly viscous until you reach about 1700 degrees Celsius. No one glazes windows that way. The fix is to add a network modifier — an oxide like soda (Na2O) or lime (CaO) whose cation refuses to join the framework. The Na+ sits in a cavity of the network as a spectator, and the oxygen it brought does something drastic: it breaks a bridge. One Si-O-Si link is cut into two dangling Si-O ends, converting a single bridging oxygen into two non-bridging oxygens, each carrying a negative charge that the nearby Na+ balances.
A GLASS NETWORK, DRAWN FLAT ( o = oxygen, * = former cation Si )
CRYSTAL (periodic) GLASS / CRN (same units, aperiodic)
o---*---o---*---o o---*---o *---o
| | | \ \ /
* o * o * * o *---o---* o
| | | / \ \
o---*---o---*---o o *---o---*---o
equal rings, repeats rings of many sizes; the angle
forever = long-range order at each shared o varies at random
a BRIDGING oxygen links two cations, forming the backbone:
...*---o---*... <- one bridge
add a MODIFIER (Na2O) and it SNIPS the bridge in two:
...*---o---*... + Na2O --> ...*---o(-) (+)Na Na(+) (-)o---*...
1 bridging oxygen two NON-BRIDGING oxygens,
charge-balanced by the Na+ ionsEach cut loosens the framework. The more non-bridging oxygens you create, the more the network is chopped into shorter, freer fragments, so its viscosity and its softening temperature both drop — which is precisely why everyday window and bottle glass is soda-lime silica (roughly 72 percent SiO2, 14 percent Na2O, 10 percent CaO by weight): cheap to melt and easy to work at a bit over 1000 degrees. But there is a limit, and it is an honest one. Add too much modifier and you snip so many bridges that no continuous network survives to freeze in — the melt simply crystallizes on cooling. A glass needs enough former to keep the framework connected. (Some oxides, notably Al2O3, are intermediates: they can slot into the network as a former or sit outside it as a modifier depending on the company they keep — a useful reminder that these categories are roles, not fixed labels.)
A glass is a liquid that stopped flowing
We have the structure; now, where did it come from? Every glass is a frozen liquid, and the freezing is unlike the one you know. Cool an ordinary liquid slowly and at its melting point it crystallizes — a sharp, first-order event where the atoms suddenly lock into a lattice and the volume drops with a distinct kink. Crystallization needs time, though: atoms must find their assigned lattice sites. Cool fast enough and the liquid slips past its melting point without crystallizing and becomes a supercooled liquid, still a disordered network but growing more and more viscous as it chills.
Keep cooling and something quietly decisive happens at the glass transition temperature Tg. The liquid has become so viscous that its atoms can no longer rearrange within the time you are giving them; the structure they happen to be in at that instant simply freezes, and the material is now a rigid solid glass. The clean way to picture it is free volume — the little pockets of extra room that let atoms shuffle past each other. As the liquid cools, that free volume shrinks; at Tg the atoms run out of room to move faster than they run out of temperature, and the arrangement is locked in. It is musical chairs: the music (thermal motion) fades, and every atom freezes wherever it stood.
Why metals fight it: quenching and frustration
Silica forms a glass almost too easily: its strong, directional covalent bonds make the melt so sluggish that the atoms can barely find the crystal even if they wanted to. Metals are the opposite extreme. Their bonds are non-directional, the atoms behave like smooth hard spheres that snap into a close-packed lattice the instant they get the chance, and an ordinary metallic melt crystallizes almost the moment you stop stirring it. To trap a metal as a metallic glass you must outrun crystallization by brute speed — rapid quenching at something like 10^5 to 10^6 kelvin per second, achieved by splatting a thread of molten alloy onto a spinning cold copper wheel (melt spinning). Freeze it before the atoms can organize, and you keep the liquid's disordered packing as a solid.
So why do some alloys form glasses at leisurely rates while others demand a million degrees a second? A large part of the answer is a beautiful idea called frustration. In a dense liquid, the packing each atom locally prefers — the lowest-energy way to surround one atom with twelve others — turns out to be an icosahedron, a cluster shot through with five-fold symmetry. But recall the crystallographic restriction theorem: no periodic crystal can have five-fold symmetry, so this locally favoured, five-fold packing simply cannot tile all of space. The liquid is frustrated — the arrangement it wants everywhere is geometrically forbidden from repeating — so it stalls between its preferred local order and any possible crystal, and nucleation becomes slow. Multi-component alloys sharpen this into the confusion principle: mix several atom sizes and no single crystal structure can satisfy them all, so even a gentle quench lands you in a bulk metallic glass.
That closes the loop on this rung nicely. A glass is a solid built from the crystal's own building blocks, keeping their short-range order and even some medium-range order, but linked with just enough randomness — a floppy hinge at every bridge, a favoured local packing that cannot repeat — that long-range order never takes hold, then frozen in place at the glass transition before it could crystallize. The covalent case is the continuous random network you have just built; the metallic case is a dense jumble of hard spheres. Guide 5 takes that metallic picture apart in full: random close packing, the geometry that lets spheres jam at about 64 percent density without ever forming a lattice.