The Physics of Glass: The Liquid That Forgot to Crystallise

The Stuff You See Through (Without Thinking About)

You’re probably looking through glass right now. The screen on your phone or laptop. The windows of whatever room you’re sitting in. Maybe your glasses, if you wear them. The glass of water on your desk — well, the glass itself is glass and the water is just water, but the English language has apparently decided that one word should do double duty.

Glass is so ordinary that it’s become invisible. Which is ironic, because invisibility is the whole point.

But here’s the thing: despite being one of the oldest manufactured materials on Earth — humans have been making glass for at least 5,000 years — we didn’t actually understand what glass is until surprisingly recently. And even now, the physics of the glass transition is considered one of the deepest unsolved problems in condensed matter physics. A Nobel laureate, Philip Anderson, once called it “the deepest and most interesting unsolved problem in solid-state theory.”

A material you can buy for a few euros per square metre, and physics still can’t fully explain it. I love that.

What Glass Isn’t

Let’s start by destroying some misconceptions, because glass has accumulated more than its fair share.

Glass is not a liquid. This is the big one — the myth that refuses to die. You’ve probably heard it: “Glass is actually a very slow-flowing liquid. That’s why old cathedral windows are thicker at the bottom — the glass has been slowly flowing downward under gravity for centuries.”

It’s a wonderful story. It’s intuitive. It sounds scientific. And it’s completely wrong.

Old cathedral windows are thicker at the bottom because of how they were made, not because they flowed. Medieval glassmakers used a technique called the crown glass process: they’d blow a large bubble of molten glass, then spin it rapidly into a flat disc. The resulting pane was inevitably uneven — thicker at the edges, thinner in the centre. When glaziers installed these panes in window frames, they typically placed the thicker edge at the bottom for stability. Some old windows are thicker at the top or on one side, which rather spoils the “flowing downward” narrative.

And the numbers seal it. At room temperature, the viscosity of window glass is approximately 10⁴⁰ Pa·s. For comparison, water’s viscosity is about 10⁻³ Pa·s. Glass at room temperature is about 10⁴³ times more viscous than water. At that viscosity, glass would need roughly 10³² years — trillions of times longer than the age of the universe — to flow by the width of a single atom. Medieval cathedrals are old. They’re not that old.

Glass is not a crystal. Crystals have their atoms or molecules arranged in a periodic, repeating lattice — neat rows and columns extending in three dimensions. Table salt, diamond, quartz, ice — these are all crystals. Glass has no such order. Its molecules are arranged in a disordered, apparently random pattern. If you could zoom into the structure of window glass at the atomic level, you’d see silicon and oxygen atoms connected in a continuous network of tetrahedra, but without any repeating pattern. It looks, at first glance, like a snapshot of a liquid.

So glass has the structure of a liquid but the mechanical behaviour of a solid. Which is exactly why it confused physicists for so long.

The Glass Transition: Freezing Without Ordering

To understand glass, you need to understand what happens when you cool a liquid.

Most liquids, cooled slowly, eventually crystallise. The molecules find their lowest-energy arrangement — a periodic crystal lattice — and lock into place. This happens at a sharp, well-defined temperature (the melting point), releases latent heat, and involves a discontinuous change in volume and entropy. It’s a first-order phase transition, clean and dramatic.

But some liquids, cooled quickly enough, miss the crystallisation window. The molecules get more sluggish as the temperature drops, moving more slowly, rearranging less. The viscosity climbs — from honey-like, to tar-like, to something essentially immovable. At some point, the molecules are moving so slowly that they can’t find the crystalline arrangement even though it’s thermodynamically favoured. They’re stuck in whatever disordered configuration they happened to be in.

That’s the glass transition. It’s not a phase transition in the traditional sense — there’s no sharp temperature, no latent heat, no discontinuous change in structure. It’s more of a kinetic arrest. The molecules didn’t choose to be disordered. They just ran out of time.

The glass transition temperature, Tg, is conventionally defined as the temperature where the viscosity reaches about 10¹² Pa·s (roughly the consistency of a material that deforms under its own weight over minutes). For soda-lime glass (ordinary window glass), Tg is around 550 °C. For pure silica glass, it’s about 1,200 °C. For polymers like polystyrene, it’s around 100 °C. For some metallic glasses, it can be below room temperature.

Here’s what makes the glass transition intellectually troublesome: Tg depends on how fast you cool. Cool more slowly, and the molecules have more time to rearrange, and Tg shifts downward. Cool faster, and Tg shifts upward. A “phase transition” whose temperature depends on your patience is not really a phase transition at all. It’s a process — a competition between cooling rate and molecular mobility — and that makes it fundamentally different from crystallisation, boiling, or any of the neat phase transitions you learn about in introductory thermodynamics.

This ambiguity bothers theorists enormously. After decades of work, there’s still no universally accepted theory of the glass transition. Is there a hidden thermodynamic transition lurking underneath the kinetics? Does an “ideal glass” exist at some temperature below Tg? Why does viscosity increase so dramatically as Tg is approached — often by 15 orders of magnitude over a temperature range of just a few hundred degrees? These questions remain open, which is remarkable for a material we’ve been making since the Bronze Age.

Why Glass Is Transparent

Here’s a question that seems simple: why can you see through glass?

Most solid materials are opaque. Metals reflect light. Wood, stone, and most plastics absorb it. Transparency is the exception, not the rule. So what makes glass special?

The answer involves how photons interact with electrons. When a photon hits a material, it can be absorbed if its energy matches the energy needed to excite an electron to a higher state. In metals, there are plenty of free electrons that can absorb photons at virtually any visible wavelength — that’s why metals are opaque (and shiny, because they also re-emit the absorbed energy).

In glass, the electrons are tightly bound in covalent bonds between silicon and oxygen atoms. The energy gap between the occupied electronic states and the next available states — the band gap — is about 9 electron volts for silicon dioxide glass. Visible light photons carry energies between 1.8 eV (red) and 3.1 eV (violet). They simply don’t have enough energy to excite electrons across the gap. They pass through without interacting.

Glass does absorb ultraviolet light, because UV photons (above about 4 eV) have enough energy to bridge the gap. This is why you don’t get sunburned through a closed car window — the glass blocks the UV-B radiation that causes sunburn. And glass absorbs some infrared wavelengths too, because the silicon-oxygen bonds vibrate at frequencies that match certain infrared photons.

But what about the lack of crystalline order? In a perfectly regular crystal, you might expect light to be scattered by diffraction — the way X-rays are scattered by crystal lattices. And indeed, transparency in crystals requires that the lattice spacing be much smaller than the wavelength of light, which it always is for visible light (atomic spacings are tenths of nanometres, visible wavelengths are hundreds of nanometres). Glass doesn’t have a lattice at all, so there’s nothing periodic to cause diffraction. The disorder is actually an advantage here — it eliminates one possible mechanism of light scattering.

There’s an additional subtlety. For a glass to be transparent, it must also be free of internal structures — bubbles, inclusions, grain boundaries, crystallites — that could scatter light. When glass devitrifies (partially crystallises), it becomes cloudy because the crystalline regions have a different refractive index from the surrounding glass, creating scattering centres. This is why glassmakers historically went to great lengths to ensure uniform cooling — too slow, and the glass might start to crystallise and lose its transparency.

I think there’s something poetic about the physics here. Glass is transparent precisely because it’s disordered. A perfect crystal of the same material (quartz) is also transparent, but for slightly different reasons. Disorder and perfect order both allow light through. It’s the messy middle — partial crystallisation, grain boundaries, impurities — that blocks it.

Silicon Dioxide: The Molecule That Makes It All Work

Window glass is mostly silicon dioxide, SiO₂, the same compound that makes up quartz, sand, and about 59% of the Earth’s crust. But the glass form of SiO₂ is structurally different from the crystalline form.

In crystalline quartz, every silicon atom is bonded to four oxygen atoms in a perfect tetrahedron, and these tetrahedra are linked in a repeating three-dimensional pattern with long-range order. In glassy silica, the tetrahedra are still there — the local bonding is identical — but the way they connect has no long-range pattern. The bond angles between tetrahedra vary slightly (instead of being fixed at the crystalline value), and the network twists and turns randomly.

This random network model was proposed by William Zachariasen in 1932, and it remains the basic picture of glass structure. The key insight is that the short-range order (each silicon bonded tetrahedrally to four oxygens) is preserved, but the long-range order is destroyed. You could examine any single silicon atom in glass and not tell whether it’s in a glass or a crystal. You’d need to look at the arrangement of hundreds or thousands of atoms to see the difference.

Pure silica glass (fused quartz) is excellent stuff — it has a very high Tg (around 1,200 °C), superb optical clarity, low thermal expansion, and high chemical resistance. But it’s extremely difficult to work with because it requires temperatures above 2,000 °C to melt. So for about 4,000 years, glassmakers have been adding fluxes — sodium oxide (Na₂O) and calcium oxide (CaO) — to lower the working temperature.

The sodium breaks some of the silicon-oxygen bridges in the network, creating what are called non-bridging oxygens. This loosens the structure, reduces viscosity at any given temperature, and lowers both the melting point and the glass transition temperature. The result is soda-lime glass — approximately 73% SiO₂, 14% Na₂O, 9% CaO, with small amounts of other oxides — which softens around 700 °C and can be blown, moulded, and shaped using relatively simple furnaces.

The trade-off? Soda-lime glass has lower chemical resistance (sodium makes it slightly soluble in water over long periods), higher thermal expansion (it cracks easily under thermal shock), and slightly lower optical quality. For everyday windows, bottles, and tableware, these compromises are fine. For laboratory glassware, telescope mirrors, and optical fibres, you need something better.

Speciality Glasses: Engineering the Impossible

The beauty of glass as a material is that you can tune its properties by changing its composition. This has produced an extraordinary family of specialised glasses:

Borosilicate glass (Pyrex, Schott Duran) replaces some of the sodium with boron oxide (B₂O₃). The result is a glass with much lower thermal expansion — roughly one-third that of soda-lime glass — which makes it resistant to thermal shock. You can pour boiling water into a cold borosilicate beaker without cracking it. Laboratory glassware, baking dishes, and high-quality scientific instruments are made from this.

Lead glass (crystal glass) contains lead oxide (PbO), sometimes up to 30%. Lead increases the refractive index, giving the glass more sparkle and brilliance — that’s why “crystal” wine glasses glitter more than ordinary ones. It also makes the glass softer and easier to cut and engrave. The high density gives it a satisfying weight. And lead glass absorbs X-rays effectively, which is why it’s used for radiation shielding in medical and nuclear facilities.

Aluminosilicate glass (Gorilla Glass, used on most smartphone screens) incorporates aluminium oxide and is then chemically strengthened by immersing it in a bath of molten potassium salt. The larger potassium ions replace smaller sodium ions in the glass surface, creating a compressive stress layer about 50 micrometres deep. This makes the glass remarkably resistant to scratching and impact — a thin sheet can be bent surprisingly far before breaking.

Optical fibres use ultra-pure silica glass doped with traces of germanium to create a core with a slightly higher refractive index than the surrounding cladding. The purity requirements are astonishing: impurities must be below parts per billion, because even trace amounts of transition metals would absorb light over the kilometres that signals must travel through fibre optic cables. Modern optical fibre is so transparent that if the ocean were made of it, you could see the bottom from the surface.

Metallic glasses are a different beast entirely. Created by cooling molten metal alloys at extreme rates (typically millions of degrees per second), they have no crystal structure at all. Metallic glasses are typically two to three times stronger than their crystalline counterparts, highly elastic, corrosion-resistant, and can be moulded like plastics. They show up in transformer cores (where their low hysteresis losses save energy), surgical instruments, golf club heads, and phone cases.

The diversity is astonishing. Same fundamental physics — a disordered, non-crystalline solid — but the properties span an enormous range depending on what atoms you use and how you arrange them.

Tempered Glass: Strength Through Stress

Ordinary glass has an annoying mechanical property: it’s strong in compression but catastrophically weak in tension. Glass fails not because the silicon-oxygen bonds are weak (they’re actually very strong — about 452 kJ/mol) but because the surface inevitably contains tiny flaws — microscopic cracks and scratches — that concentrate stress. When the glass is pulled apart (put under tension), these cracks propagate at the speed of sound, and the glass shatters.

Tempering solves this by putting the surface of the glass into permanent compression. The process is simple in concept: heat the glass to about 620 °C (above Tg but below the melting point), then blast both surfaces with cold air. The surfaces cool and solidify first, while the interior is still hot and soft. As the interior eventually cools and contracts, it pulls the already-rigid surfaces inward, putting them into compression.

Now, for a crack to propagate through the surface, the applied tensile stress must first overcome the built-in compressive stress. Tempered glass is typically four to five times stronger than annealed glass of the same thickness. This is why car side windows, shower doors, and glass tables use tempered glass.

There’s a dramatic consequence of this stored stress. When tempered glass does eventually break — if something punches through the compressive layer — the entire sheet shatters simultaneously into small, relatively harmless granules rather than large, dangerous shards. The stored elastic energy releases all at once, and the entire glass explodes into fragments. If you’ve ever seen a car window break, you’ve seen this: one crack, and suddenly the entire pane is a mosaic of pebble-sized pieces.

Prince Rupert’s drops demonstrate this principle in extreme form. Drop molten glass into cold water, and the rapid quenching creates a tadpole-shaped droplet with enormous compressive stress in the surface. The bulbous head is nearly indestructible — you can hit it with a hammer and it survives. But snap the thin tail, and the entire drop explodes violently into powder. The stress imbalance propagates as a crack wave travelling at about 1,500 metres per second, shattering the entire structure in microseconds. High-speed cameras have captured this beautifully, and I’d recommend looking it up if you’ve never seen it.

Glass in Nature

Humans didn’t invent glass — nature got there first, several times over.

Obsidian is natural volcanic glass, formed when silica-rich lava cools so rapidly that crystals don’t have time to form. It’s been used for cutting tools since the Stone Age because it fractures conchoidally — breaking along smooth, curved surfaces that produce edges sharper than any steel blade. Obsidian surgical scalpels, with edge widths measured in nanometres, have been used in experimental eye surgery because they cause less tissue damage than steel.

Fulgurites are tubes of glass formed when lightning strikes sand. The bolt superheats the sand to above 1,800 °C in milliseconds, fusing the silica grains into a hollow, branching tube of glass. They’re geological fossils of individual lightning strikes, sometimes extending metres underground.

Tektites are glass beads formed from terrestrial rock melted and ejected by large meteorite impacts. They’re found in strewn fields hundreds of kilometres long, and their composition matches local surface rocks, confirming their terrestrial (not extraterrestrial) origin.

And then there’s something stranger: water glass. Water can form a glass if cooled quickly enough — below about −137 °C without crystallising. Vitrified water is thought to be the most abundant form of water in the universe, coating dust grains in interstellar space and making up much of the ice on comets. When the Rosetta mission analysed comet 67P, it found evidence of amorphous (glassy) water ice — water that formed in the cold of deep space and never had the thermal energy to arrange itself into crystalline ice.

The idea that the most common form of water in the cosmos might be glass — not the crystalline ice we know on Earth — is one of those facts that makes me stop and reconsider how parochial our everyday experience really is.

The Unsolved Problem

I promised you an unsolved problem, and here it is.

As a liquid approaches its glass transition temperature, its viscosity increases dramatically. For some glass-forming liquids (called “fragile” liquids, in a technical classification that has nothing to do with brittleness), the viscosity increases by 15 orders of magnitude over a temperature decrease of just 200–300 degrees. This is far steeper than simple thermal physics would predict.

Why? What is happening at the molecular level that makes relaxation times increase so explosively?

One popular idea is the Adam-Gibbs theory: as the liquid cools, the number of possible configurations the molecules can explore decreases. Eventually, the molecules must rearrange cooperatively — in groups — rather than individually. These cooperatively rearranging regions grow in size as the temperature drops, and larger regions take exponentially longer to rearrange. The viscosity blows up because the molecules literally can’t find their way to new configurations fast enough.

Another approach — the mode-coupling theory — treats the slowing down as a kind of self-generated cage effect. Each molecule is trapped in a cage of its neighbours, and those neighbours are trapped in cages of their neighbours, and so on. At a critical temperature, the cages become permanent, and the liquid freezes into a glass.

A third perspective proposes that a true thermodynamic phase transition to an “ideal glass” exists at some temperature below Tg — the Kauzmann temperature — but we can never reach it experimentally because the system falls out of equilibrium first. The glass transition we observe is just the kinetic shadow of this hidden thermodynamic transition.

Each theory captures some aspects of the data but fails on others. Decades of experiments, simulations, and theoretical work have not produced a consensus. The glass transition remains what Anderson called it — one of the deepest unsolved problems in condensed matter physics.

And honestly? I think that’s wonderful. A material so common you’re probably touching it right now, and we still can’t fully explain how it forms. It’s a good reminder that “everyday” and “understood” are not synonyms.

What Glass Teaches Us About Order and Disorder

Glass sits at the boundary between order and chaos. It’s not the neat perfection of a crystal, with every atom in its place. It’s not the randomness of a gas, with atoms flying in every direction. It’s something in between — a structure that looks random but has hidden correlations, that’s frozen in time but not in equilibrium, that’s thermodynamically unstable but kinetically permanent.

In a way, glass is a monument to the importance of history. Two pieces of glass with identical compositions can have different properties depending on how they were cooled. The structure remembers the cooling process. Fast cooling traps a higher-energy, more disordered structure. Slow cooling allows more relaxation, producing a denser, lower-energy glass. The material carries a memory of its thermal history in its atomic arrangement.

This sensitivity to process — to path rather than just destination — makes glass fundamentally different from crystals. A crystal of quartz is a crystal of quartz, regardless of how it was made. But a glass of silica can be many different things, depending on the details of its creation.

I think there’s a metaphor in there somewhere, but I’ll resist the urge to stretch it too far. The physics is interesting enough without making it carry philosophical weight.

Five thousand years of glassmaking. Billions of windows, lenses, bottles, screens, and optical fibres. And the fundamental question — what is this stuff, really? — still doesn’t have a complete answer.

That’s the kind of problem that keeps physics interesting.