Fibre strength and flaws
A Socratic walk-through of fibre strength and flaws — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why is a hair-thin glass fibre far stronger than a thick rod of the very same glass?
Strength, as we normally use the word, is a property of a material. Steel is stronger than pine, and that statement need not say how thick the piece is, because we quote strength as force per unit area precisely so that thickness cancels.
So a fibre drawn from a melt should have the same strength per unit area as a rod poured from the same melt. It does not; the fibre can carry a stress a hundred times greater. Something in the framing is wrong, and the interesting question is which part.
Reasoning it through
REASONING #Start by asking what sets the strength of ordinary glass, because that is the anomaly, not the fibre. Silica's bonds are strong; from bond energies you would predict a material capable of enormous stress. Window glass instead fails at a few tens of megapascals — and, more damning, at unpredictable stresses, so two panes off the same line can differ severalfold. A number that is a property of the chemistry does not scatter like that.
What varies between two panes is not chemistry but damage. Every glass surface carries invisible flaws: handling scratches, contact points, nicks. Griffith's insight, a century ago now, was that a crack concentrates stress at its tip in proportion to the crack's depth — so strength falls roughly as the inverse square root of the worst flaw present. Glass at room temperature has no way to blunt a crack tip: no plastic flow, no dislocation motion. Once the tip moves, nothing arrests it.
So strength is set by the single worst flaw, not by the average condition of the material, and a body has no strength of its own — only the strength of its weakest link.
The size effect then follows almost for free. If failure is governed by the worst flaw in the stressed region, a specimen that stresses less material samples fewer flaws, and the worst of a small sample is typically milder than the worst of a large one. That is not a claim about the material; it is extreme-value statistics, and it predicts something beyond the diameter effect: a longer fibre should be weaker than a short one of the same diameter. It is, measurably, and that length dependence is routine in optical-fibre testing. The prediction convinces me the statistics are real, because thickness plays no part in it.
But are the statistics the whole story? Let me try to break my own account. If the fibre's advantage were purely that it samples fewer flaws, two things should hold: a thin fibre that has been rubbed against something should still be fairly strong, and a thick rod whose surface has been etched away should still be weak. Neither does. A drawn fibre touched once by a hard object loses most of its strength immediately, which is why optical fibre is coated in-line within centimetres of the draw furnace and never allowed to touch anything bare. And glass rods etched in hydrofluoric acid, which strips the damaged surface layer, become dramatically stronger while remaining thick.
So the statistical size effect is real but secondary. The dominant reason a fibre is strong is that its surface is new — formed by drawing from a melt, never abraded, protected before anything could reach it. Thinness helps further by limiting how deep a flaw can geometrically be. Griffith's own extrapolation, that strength would rise without limit as diameter fell toward zero, is now generally read as an artefact of how his specimens were made and gripped rather than a law about diameter.
Which reframes the engineering. You do not get strong glass by finding better glass; you get it by making surfaces and never touching them — and, since service cannot guarantee that, by designing around flaws you must assume are there. Telecommunications fibre is proof-tested by pulling every metre well above its service load, so anything seriously flawed breaks in the factory.
The analogy
THE ANALOGY #Think of a chain hauled up from the sea floor, every link corroded at its own rate. The chain's capacity is that of its worst link, and nobody can say in advance which link that is. A short length contains fewer links, so its worst is likely better than the worst in a long one — but that is a weak advantage beside the real one, which is having a chain forged this morning that has not yet been in the sea.
the links of a chain fail one at a time and you can replace the bad one, whereas a crack in glass runs through the whole body in a single event, and no part of the piece survives to be repaired.
Clarifying the model
THE MODEL #This is a property of brittle materials specifically. A ductile metal blunts a crack tip by flowing plastically around it — the dislocation motion the alloy-strength explanation in this collection is about — so metals show a far weaker size effect. Brittleness and weakest-link behaviour are the same fact seen from two sides.
Strength here is also a distribution rather than a number: the right question about a batch of fibre is not "how strong is it" but "what fraction fails below the design stress".
One sibling explanation is worth marking the boundary against. Tempered glass also gets its strength from flaws, but by the opposite move: it leaves them in place and holds them shut with locked-in surface compression, so a load must cancel that compression before any crack tip feels a pull. That neutralises flaws you cannot remove; the fibre's way is never to acquire them. The two can be combined, and chemically strengthened glass does both.
An honest caveat about numbers. Reported pristine-fibre strengths depend heavily on gauge length, test method, and how fast the test runs, because glass also fails slowly under sustained load in the presence of moisture. A single "strength of silica fibre" is a simplification of a family of curves.
A picture of it
THE PICTURE #How to readTwo independent contributions, one per axis: rightward means less material under stress and so fewer flaws sampled, upward means a less damaged surface. Read the vertical spread first — etched rod sits high and plate glass low though both are bulky, showing surface condition alone moving the result a long way. Then read horizontally: abraded fibre is as thin as drawn fibre and still weak, which is the test that rules out thinness as the main cause. Placement is qualitative judgement about the two conditions, not measured strengths.
What became clearer
WHAT CLEARED #The fibre is not made of better glass; what has to give is the framing that strength is a property of a material. In a brittle solid, strength is a property of the worst flaw in the stressed region — so it belongs to the specimen, its history and its size, not to the substance. A fibre wins mostly because its surface is freshly formed and was protected before anything could scratch it, and only secondarily because a thin, short specimen samples a smaller population of flaws. Touch it once and the advantage is gone, which tells you which of the two mattered.
Where to go next
ONWARD #- How glass fibres in a resin matrix let a composite exploit fibre-scale strength at structural scale.
- Why glass under sustained load fails hours or years later, through slow crack growth assisted by water.
Key terms
TERMS #| Term | What it means |
|---|---|
| Griffith criterion | the condition for a crack to propagate, giving strength falling roughly as the inverse square root of flaw depth. |
| Weakest-link statistics | treating a body's strength as the minimum over the flaws it contains, from which the size effect follows. |
| Proof testing | deliberately loading every unit above its service stress so flawed units fail before delivery. |
| Static fatigue | delayed failure of glass under sustained load, driven by moisture-assisted slow crack growth. |
Every term the collection defines is gathered in the glossary.