Choosing a Material: The Ashby Approach

From "what is it?" to "which one?"

Everything below this rung ran the arrow one way. You learned that a material's structure sets its properties, that processing shapes the structure, and that properties add up to performance — the whole structure-property-processing-performance chain, read left to right. This capstone rung runs the arrow backwards. A part has a job to do; you must reach into a menu of well over a hundred thousand materials and pull out the right one. That is materials selection, and doing it by hunch — "it carries a load, use steel" — is exactly how good engineers get blindsided.

The reason hunches fail is that the words we use to praise a material are not one word. "Good" splinters into properties that do not travel together. Stiffness (how much it springs back, set by Young's modulus) is not strength (the stress it yields at) is not toughness (how much energy it soaks up before it snaps). A ceramic is enormously stiff and strong yet shatters like glass; annealed copper is tough and formable yet embarrassingly soft; foam is featherlight yet floppy. Ask only "what is strong?" and you will happily bolt a heavy steel panel where a lighter aluminium or carbon-fibre one would have been stiffer per kilogram. The first discipline of selection is to name, precisely, which property actually matters for this part.

Translating the design: function, constraints, objective, free variables

Ashby's core move is a translation. Before you look at a single material, you rewrite the vague design brief into four sharp lines. Get the translation right and the rest is almost mechanical; get it wrong and no amount of clever ranking can save you. The four lines are the same for a bicycle frame, a spring, a turbine disc or a phone case.

1. Function — what the part must DO. Carry a tensile load? Store elastic energy? Conduct heat away? A tie carries tension; a beam resists bending; a shaft carries torque. The function fixes which equations govern the part.
2. Constraints — the non-negotiables it must satisfy. It must not yield or fracture; it must fit a given length; it must survive 300 degrees C; it must not corrode in seawater. These are pass/fail gates, not things to optimise.
3. Objective — the single thing you want to make extreme. Almost always: minimise mass, or minimise cost — occasionally maximise stored energy or minimise environmental impact. This is what you push to a maximum or minimum.
4. Free variables — what you are free to choose. Always the material itself; usually also a dimension you have not yet fixed, such as the cross-sectional area or a wall thickness. The art is to eliminate these using the constraints.

Make it concrete with the simplest possible part: a straight tie-rod of a fixed length L that must carry a tensile load without stretching more than a set amount — a light, stiff tension member, the kind hiding inside every truss and bike frame. Function: a tie in tension. Constraint: it must be stiff enough, meaning its stiffness S = A x E / L must hit a target value (A is the cross-section, E the Young's modulus). Objective: minimise the mass m = A x L x rho. Free variables: the cross-section A (we can make the rod fatter or thinner) and the material (E and rho). Two unknowns, one constraint — and that is exactly enough to squeeze out a clean answer.

The performance index: why E over rho

Now do the algebra that turns a design into a material score. The constraint fixes the stiffness: S = A x E / L, so the cross-section we are forced to use is A = S x L / E. Substitute that into the mass we want to minimise: m = A x L x rho = (S x L / E) x L x rho = S x L^2 x (rho / E). The design targets S and L are fixed by the customer; the only part of that expression the material controls is the last bracket, rho / E. To make the rod as light as possible we minimise rho / E — which is the same as maximising E / rho, the material's specific stiffness or stiffness-to-weight ratio. That combination, E / rho, is the performance index for a light stiff tie, and the material with the biggest value of it wins. Notice what happened: the loads, the length and the geometry all cancelled out, leaving a pure property group. That is the whole magic of the method.

Put real numbers on it and a famous surprise appears. Steel: E about 200 GPa, rho about 7.8 g/cm^3, so E / rho is roughly 26. Aluminium: E about 70 GPa, rho about 2.7, so E / rho is again roughly 26. They tie exactly — a stiff aluminium tie and a stiff steel tie of the same stiffness weigh the same, which is why swapping steel for aluminium buys you nothing on a pure tension member. Wood along the grain (E about 10 GPa, rho about 0.5) scores around 20, astonishingly close to both metals. Carbon-fibre composite (E about 130 GPa, rho about 1.6) scores about 80 and walks away with it. "Use the strongest metal" would never have found that ranking.

The Ashby chart: seeing the answer

You could compute E / rho for ten thousand materials and sort a spreadsheet, but Ashby's real gift is a picture. Plot Young's modulus up the page against density across it, both on logarithmic axes, and every material family gathers into a bubble — a blob, not a point, because each family spans a range. Metals sit dense and stiff on the right; technical ceramics ride the top; polymers cluster low and middling; foams and elastomers slump into the light, floppy bottom-left; and composites straddle a coveted middle-upper patch, light yet stiff. The whole world of materials fits on one material property chart, and the empty white regions are just as telling — they are combinations nature and industry have not yet delivered, which is where the frontier lives.

  E (GPa)   MATERIAL PROPERTY (ASHBY) CHART: stiffness vs density (log-log)
  ^
1000|                                    .-----------.
    |                                    (  CERAMICS  )
 100|                     .-----------.   '-----------'
    |                     ( COMPOSITES )  .-----------.
  10|         .-----.     '-----------'   (   METALS  )
    |         ( WOOD )                    '-----------'
   1|         '-----'      .-----------.
    |                      (  POLYMERS  )
 0.1|        .-------.     '-----------'
    |        ( FOAMS  )  .-------------.
0.01|        '-------'   ( ELASTOMERS  )
    |                    '-------------'
    +--------------------------------------------------> rho (g/cm^3)
     0.03    0.1    0.3      1        3       10

  A performance index E/rho = C is a STRAIGHT LINE of slope 1 on these
  log axes. Slide that line toward the TOP-LEFT (stiffer AND lighter);
  the last bubble it touches is the winner. Steeper lines (slope 2, slope 3)
  are the beam index E^(1/2)/rho and the panel index E^(1/3)/rho.

This is where the log axes stop being a formatting choice and become the whole trick. Because E / rho = C turns into the straight line log E = log rho + log C, every performance index is a ruler of known slope: the tie index E / rho is slope 1, the beam index E^(1/2) / rho is slope 2, the panel index E^(1/3) / rho is slope 3. Lay the correct ruler on the chart and push it toward the top-left corner — stiffer and lighter at once — and the last bubble it kisses is the lightest material that does the job. You have turned a ten-thousand-row search into sliding a line across a single page, and you can literally see why the winner wins.

Process, cost, judgment — and the road ahead

An index hands you a shortlist, never a final answer, and pretending otherwise is where selection goes wrong. A material you cannot shape, join or afford is no material at all. Material and process are married: carbon fibre needs slow, skilled layup; a fiddly bracket may demand a shape you can only cast or 3D-print; some superb alloys cannot be welded. And cost is often the loudest voice in the room — in mass production the cheapest material that clears every constraint usually wins, which is why humble steel and concrete still build most of the world despite losing many an index. Real cost and availability can overrule a beautiful stiffness-to-weight score in a single line of a bill of materials.

Be honest about what the method is and is not. Each index is a first screen built on idealisations — one objective, one loading mode, linear-elastic behaviour, a single dominant property. Real parts juggle several objectives that actively fight each other: light AND cheap AND recyclable rarely point to the same material, so you end up trading them off along a Pareto front, where improving one can only cost you another. Multiple constraints stack up; safety factors and worst-case flaws lurk; the humble questions of manufacturability and supply refuse to fit on the chart. The charts and indices do not replace engineering judgment — they organise it, so your judgment is spent on the genuinely hard trade-offs instead of on brute-force searching.

That maps the rest of this capstone. Guide 2 builds the property charts and performance indices in full, deriving the beam and panel exponents you were just handed. Guide 3 widens the objective from mass to money and to the planet — manufacturing cost, life-cycle assessment, embodied energy, recycling and the circular economy, so that "which material?" also asks "at what cost to the world?" Guide 4 turns selection inside out into detective work: failure analysis, reading a broken part to learn what its material could not do and why it was the wrong pick. And guide 5 leaves the existing menu behind for the frontier, where we no longer just choose materials but invent them — graphene and carbon nanotubes, biomaterials for implants, the batteries and solar cells of energy storage, shape-memory and metamaterials, and computational materials science that hunts new compounds by machine learning before a single crucible is lit. Selection is where all the science you climbed becomes a decision; the frontier is where the decisions run out of materials and we make more.