Break the same glass ten times and get ten different strengths. That is not sloppy measurement — it is the material

Look up the yield strength of mild steel and you get a number you can design with. Look up the strength of glass and you get a range so wide it is nearly useless — and if you test twenty identical glass rods yourself, you will get twenty different breaking loads, scattered by a factor of two or more.

That scatter is not poor technique. It is the honest answer, because a brittle material does not really have a strength. It has a worst flaw, and the strength you measure is a statement about that flaw rather than about glass.

Once that is clear, a set of otherwise baffling everyday facts line up: why a phone screen survives a drop onto concrete and then dies on a gentle knock at the corner, why toughened glass disintegrates into harmless cubes instead of shards, why you cannot cut it afterwards, and why a shelf that held its load all evening can let go at two in the morning.

Strength is decided at the tip of the worst crack

Calculate how strong glass ought to be from the strength of the silicon–oxygen bonds holding it together, and you get something in the region of several gigapascals. Measured glass usually breaks somewhere around fifty to a hundred megapascals — one or two per cent of the theoretical value.

The explanation, worked out by A. A. Griffith in the 1920s, is that stress does not distribute itself evenly through a flawed solid. At the tip of a sharp crack, stress concentrates enormously, and the sharper and longer the crack, the higher the concentration. A crack propagates when the energy released by advancing it exceeds the energy needed to create the two new surfaces — which means failure is governed by the largest, sharpest, worst-oriented flaw present, not by the average condition of the material.

A ductile metal has an escape route: at a stress concentration it yields locally, deforming plastically and blunting the crack tip, spreading the load. Glass and ceramics have no such mechanism at room temperature. There is nothing to stop a crack once it starts, which is why brittle failure is sudden and complete rather than gradual.

And critically, the flaws that matter are almost always surface flaws, because that is where handling puts them. This gives rise to one of the most striking demonstrations in materials science: a freshly drawn glass fibre, before it touches anything at all, is extraordinarily strong — approaching the theoretical limit. Let it brush against a guide, a glove or a speck of dust and it loses most of that strength permanently.

That is not a laboratory curiosity. It is why optical fibre is coated with polymer within centimetres of being drawn, while still in the tower, before anything can touch the glass. The coating is not there for optics. It is there because the fibre's strength exists only as long as its surface is pristine.

Why bigger is weaker, and why it is a probability

If strength is set by the worst flaw, then two consequences follow that do not apply to metals at all.

A large piece is weaker than a small one. Not proportionally weaker — weaker in stress terms, per unit area. A bigger piece contains more surface, so it has more opportunities to contain a serious flaw, and it only takes one. This is why testing a small sample and scaling up the result is a reliable way to be wrong about a ceramic, and why size effects have to be designed for explicitly.

Strength is a distribution, not a value. The standard tool is Weibull statistics, which describes the probability of survival as a function of applied stress. A Weibull analysis does not tell an engineer what a ceramic's strength is; it tells them the stress at which, say, 99% of pieces will survive, together with a modulus describing how tightly the population is grouped. Designing with brittle materials means choosing an acceptable failure probability, which is a different mental activity from looking up a yield strength and applying a safety factor.

Then there is the fact that genuinely surprises people.

Glass gets weaker while you watch

A piece of glass that survives a load applied instantly can fail under the same load applied for long enough. This is static fatigue, and unlike metal fatigue it requires no cycling at all — just time, stress and moisture.

The mechanism is stress-corrosion at the crack tip. Water molecules reach the highly strained silicon–oxygen bonds at the tip of an existing flaw and react with them, breaking bonds that mechanical stress alone could not. The crack creeps forward, slowly, and as it lengthens the stress concentration at its tip rises, so it creeps faster. Eventually it reaches the length at which it runs, and the piece fails — hours or months after the load was applied and nothing changed.

This is why glass structural elements are designed to much lower sustained stresses than their instantaneous strength would suggest, why humidity matters in the specification, and why a cracked windscreen or a chipped shelf keeps getting worse without anybody touching it. It also means "it held yesterday" is a genuinely unsafe argument for a brittle material in a way it is not for steel.

A metal's strength is a property of the metal. A brittle material's strength is a property of the particular object in front of you, its surface history, how large it is, how long the load stays on, and how humid the room is. The number in the handbook is a starting point for a calculation about probability, not a limit you can rely on.

The fix: make the surface want to stay closed

Since failure starts at surface flaws under tension, both practical strengthening methods do the same thing — they put the surface into permanent compression, so an applied tensile load must first overcome that compression before any crack can open at all. The flaws are still there. They simply cannot be pulled apart.

Thermal tempering heats the glass and then quenches its surfaces rapidly with air. The outside solidifies first; the interior cools and contracts afterwards, pulling the already-rigid skin into compression and leaving the core in balancing tension. The result is several times stronger in bending and, when it does finally fail, the stored energy in that tensioned core shatters it into small blunt cubes rather than blades. That is why it is used for car side windows, shower screens and glass doors — the failure mode is designed to be survivable.

Two consequences follow directly. The glass cannot be cut, drilled or ground after tempering, because breaching the compressive skin releases the whole stress balance at once — all fabrication must happen before. And a deep scratch or an impact reaching past the compression layer causes immediate, total disintegration, which is why toughened panels sometimes appear to explode spontaneously, often from a flaw or inclusion at an edge.

Chemical strengthening achieves the same end differently and is what makes modern phone screens possible. The glass is bathed in a molten potassium salt, and potassium ions diffuse in to replace the smaller sodium ions already in the surface. The larger ion is now wedged into a site sized for a smaller one, and the resulting crowding puts the surface under compression. Because this is a diffusion process, it works on thin sheets that could never be thermally tempered, and it produces a much deeper and more uniform compressive layer than its thickness would suggest.

The design trade-off is worth knowing because it explains phone screen behaviour. High surface compression resists scratching and flexing; deep compression is what stops a flaw that penetrates further — a sharp impact on a hard rough surface such as concrete or a granite floor introduces damage below shallow compression. This is precisely why a phone can survive several flat drops and then fail on a modest corner impact: corners concentrate stress and the strike can push a flaw past the protective layer.

Zirconia, the ceramic that fights back

The other route is to give the ceramic something to do at a crack tip, and the most elegant example is a metal oxide from the same family as much of this catalogue.

Zirconia can exist in more than one crystal structure, and a particular one — the tetragonal phase — can be held in place at room temperature by adding a stabiliser such as yttria, even though it would prefer to be monoclinic. The stabilised phase is metastable: it is waiting for an excuse.

When a crack starts to run through the material, the intense stress field at its tip provides that excuse. Grains near the tip transform to the monoclinic phase, and in doing so they expand in volume by a few per cent. That expansion, occurring in a confined space directly around the crack tip, squeezes the crack shut and absorbs energy. The material responds to being cracked by clamping down on the crack.

This is transformation toughening, and it makes zirconia markedly tougher than ordinary ceramics — tough enough for dental crowns, cutting blades, pump components and femoral heads in hip replacements. It is one of the few genuinely clever answers to brittleness rather than a workaround for it.

The honest caveat is equally instructive. The same metastability that provides the toughening can be triggered by moisture and warmth over time without any crack at all, causing the surface to transform gradually, roughen and microcrack. This low-temperature degradation is a documented failure mechanism and was implicated in problems with a generation of zirconia hip implants, which is why modern formulations and processing are specifically designed to resist it. A material that works because it is on the edge of a transformation will always need that edge managed carefully.

Where the nanoscale comes in, and what it does not fix

Fine-grained and nanostructured ceramics are an active area, and the reasoning is sound: in many ceramics the flaw population is related to the microstructure, so a finer, more uniform grain structure can reduce the size of the largest defect and raise both strength and its consistency — that is, a higher Weibull modulus, which for a designer is often worth more than a higher average.

The limit is that grain size is not the only source of flaws. Porosity left by incomplete sintering, inclusions, agglomerates that did not break up during forming, and damage introduced during machining and handling all create defects that have nothing to do with how fine the powder was. A nanostructured ceramic made from a poorly dispersed powder inherits the agglomerates as flaws, and the finished part is only as good as its worst one.

Which returns to the theme running through this whole subject: for a brittle material, an average tells you very little and the extreme tells you everything.

Why it matters for students and researchers

Brittle materials are the clearest case in engineering where a property has to be treated statistically rather than deterministically, and learning that early changes how a student reads every datasheet afterwards. A strength figure for a ceramic without a Weibull modulus, a specimen size and a test geometry is not a specification — and unlike many such complaints, this one changes the answer by a factor of two rather than a few per cent.

It is also a good lesson in where a material's properties actually live. Glass strength is a property of the surface, created and destroyed by handling, which means the manufacturing and logistics of a ceramic component are part of its mechanical performance rather than a separate concern. The same insight that runs through adhesion, shielding and filtration applies here too: the system is governed by its worst local feature, not by its bulk average.

The open problems are practical. Non-destructive detection of critical flaws in finished ceramic parts remains limited, which is why proof testing — deliberately loading every part to a stress above its service load, so that survivors are known to be free of flaws above a certain size — is still widely used and still wasteful. Reliable lifetime prediction under sustained load and humidity is genuinely difficult. And toughening mechanisms beyond transformation toughening, including fibre reinforcement and engineered weak interfaces that deflect cracks, are where the field's more interesting recent work sits.

Frequently asked questions

Why did my phone survive a big drop and then crack on a small one?

Because what matters is not the height but where and how the impact lands, and what the surface hits. A flat landing spreads the load; a corner or edge strike concentrates it and can drive damage below the compressive layer that protects the screen. Landing on a hard, rough surface such as concrete or stone flooring introduces sharper, deeper flaws than a smooth surface does. And accumulated micro-damage from earlier drops means the screen you dropped today was not the screen you bought.

Is tempered glass stronger than ordinary glass?

Substantially, in bending, and that is its purpose. The trade-offs are that it cannot be cut or drilled after treatment, and that once failure begins it is total — the stored energy releases at once, which is deliberate, since a shower of small blunt cubes is far safer than large sharp shards. For places where the glass must stay in one piece after breaking, laminated glass with an interlayer is used instead, and car windscreens are laminated for exactly that reason.

Why can't tempered glass be cut to size?

Because its strength comes from a balanced stress state: a compressed skin held in equilibrium by a tensioned core. Cutting or drilling breaches the skin and releases that balance everywhere at once, and the panel disintegrates. All cutting, drilling and edge finishing must be completed before tempering, which is why toughened panels are made to order rather than trimmed on site.

Do screen protectors actually help?

Yes, and the reason follows directly from this article. A brittle material fails from surface flaws, so a sacrificial layer that takes the scratches and the initial impact damage keeps the screen's own surface pristine. A protector cannot prevent a severe impact from flexing the display glass past its limit, but it genuinely reduces the accumulation of the small surface defects that lower strength over time.

Why does a crack in a windscreen keep growing when nothing is touching it?

That is static fatigue. Water molecules reach the strained bonds at the crack tip and react with them, so the crack extends slowly under stresses far below those needed to break the glass outright — and as it lengthens, the stress at its tip rises and it advances faster. Temperature swings and vibration accelerate it. This is why a chip is repaired promptly rather than watched, since the repair is straightforward and the crack is not going to stop on its own.