Why Materials Are Not as Strong as They Should Be

"Strong" is used for at least three unrelated things, and conflating them is how components get specified badly.

Glass is stiffer than aluminium and far weaker in practice, because it has almost no toughness. Rubber is enormously extensible and not remotely stiff. A material can be high in any one of these and low in the others.

Gotcha: "It bent, so it was not strong enough" and "it snapped, so it was not strong enough" describe opposite failures. The first is insufficient stiffness or yield strength; the second is insufficient toughness, and the fix for one often makes the other worse.

Young's modulus is what you get when you stretch interatomic bonds slightly and they pull back. Strong, directional bonds give a high modulus; weak or entropic ones give a low one.

Steel is about four thousand times stiffer than rubber. It is also three times stiffer than aluminium, which is worth remembering the next time someone proposes saving weight by switching: at equal geometry the aluminium part deflects three times as far.

Rubber is the outlier in kind, not just degree. Its stiffness does not come from stretching bonds at all but from uncoiling tangled polymer chains, which is why rubber gets stiffer when heated while every metal gets softer.

Here is the fact that organises the whole subject. Stiffness is essentially fixed by chemistry. Strength is almost entirely a matter of microstructure.

Heat-treat a steel, quench it, hammer it, alloy it, refine its grains: you can move its yield strength by a factor of ten. Its Young's modulus stays at about 210 GPa throughout. Every steel, from a soft mild steel to a hardened tool steel, has essentially the same stiffness.

Key idea: Stiffness is structure-insensitive; strength is structure-sensitive. If a part deflects too much, no heat treatment will help and you must change material or geometry. If a part yields, you have a large menu of options that do not require changing material at all.

This also means strength is not really a property of a substance. It is a property of a particular piece of that substance, with its particular history of processing.

Plastic deformation in a metal is one plane of atoms sliding over another. So calculate what it takes to make an entire plane slide at once.

The atoms sit in a periodic potential, and the restoring force is roughly sinusoidal in the displacement. Slide them a quarter of a lattice spacing and the resistance peaks. Working through the geometry gives

$\tau_{\text{theoretical}} \approx \frac{G}{2\pi}$

where $G$ is the shear modulus. Copper has $G \approx 45$ GPa, so its perfect crystal should resist shear up to about 7 GPa.

That calculation is not subtle and it is not wrong. It was done in the early twentieth century and checked repeatedly, because the answer it gives is absurd.

This gap was the central problem of metallurgy for a generation. Nobody could see a dislocation, and their existence was proposed in 1934 by Taylor, Orowan and Polanyi independently, purely because nothing else could account for the discrepancy.

Direct observation had to wait for transmission electron microscopy in the 1950s, more than twenty years after the theory. Between those dates the entire structure of modern metallurgy was built on an entity nobody had ever observed.

Brittle materials have the same problem from the other direction, and A. A. Griffith found the clue in 1920, in his paper on the phenomena of rupture and flow in solids.

Griffith drew glass into fine fibres and measured their strength. The thinner the fibre, the stronger it was, tending toward the theoretical value as the diameter shrank. Bulk glass, made of exactly the same substance, is weak.

His explanation: every real surface carries microscopic cracks, and a crack concentrates stress at its tip. Strength is set by the largest flaw present, not by the bonds. A thinner fibre simply has less surface in which a large flaw can exist.

Key idea: Two different classes of defect, dislocations inside the crystal and cracks at the surface, both explain the same puzzle: real materials fail far below their theoretical limit because failure starts at the worst spot, not the average one. Every strengthening technique in this course is a way of managing defects, never of eliminating them.

Defects are classified by dimensionality, and each class controls different properties.

A perfect crystal is not the goal. A metal with no dislocations at all would be extremely strong and completely brittle, since dislocation motion is plastic deformation, and plastic deformation is what lets a material absorb energy instead of shattering.

In practice: The engineering objective is never to remove defects but to arrange them: enough dislocations to keep the material ductile, obstructed enough to keep it strong, and no large voids or inclusions to start a crack.

Pull a metal bar and record force against extension. The curve has four regions and every one is a design decision.

1. Elastic. Stress is proportional to strain, $\sigma = E\varepsilon$, and releasing the load returns the bar exactly to its original length. Bonds stretch and relax.
2. Yield. Dislocations start moving. Deformation becomes permanent. This is the number most designs are built around, with a safety factor on top.
3. Work hardening. The curve keeps rising past yield, because moving dislocations tangle with each other and obstruct further motion. The material gets stronger as you deform it.
4. Necking and fracture. At the ultimate tensile strength, deformation localises into one narrowing region and the load drops until the bar parts.

Gotcha: Ultimate tensile strength is the number quoted on datasheets and is almost never the one to design to. A part that has passed yield is permanently deformed and usually already failed in service, whatever the UTS says.

Strengthening a metal means obstructing dislocation motion. Ductility comes from dislocation motion. These are the same mechanism seen from two sides, so almost every strengthening treatment costs ductility.

A fully hardened steel might yield at 1,500 MPa and break after 2 percent elongation. The same steel annealed yields at 250 MPa and stretches 40 percent before parting. Which one you want depends entirely on the failure you are worried about.

Key idea: Ductility is not a weakness, it is a warning system. A ductile part deforms visibly before it fails, redistributes load away from stress concentrations, and absorbs energy in a crash. A very strong brittle part gives none of that and fails without notice.

This is why aircraft structure is aluminium and titanium alloys rather than the strongest available steel, and why bridge steel is specified with a minimum elongation as well as a minimum strength. Beating the trade-off, rather than accepting it, is what the next two lessons are about.

Everything so far reduces to one claim: a material's mechanical behaviour is set by its defects, and defects can be engineered.

• Next comes the dislocation itself: how it moves, why some crystal structures allow it and others do not, and the four standard ways of obstructing it. That is where the factor of ten in strength comes from, and where the ductility gets spent.
• Then fracture and fatigue: what happens when a crack exists rather than a dislocation, why a component survives one big load and dies of a million small ones, and the two accident investigations that created the field.
• Finally selection: how to choose among thousands of candidate materials using two or three numbers, and which failure mode to check for before committing.

In practice: If you take one thing from this lesson, take the distinction in step one. Most material specification arguments turn out, on inspection, to be two people using "strong" to mean different properties.