Material selection by properties

In brief

You are asked to design a bicycle frame that is light, stiff enough not to flex under pedaling, strong enough not to snap, and affordable enough to sell. Steel is cheap and strong but heavy. Aluminum is lighter but less stiff for the same shape. Titanium is light and stiff but expensive. Carbon fiber can beat all three on stiffness per weight but costs more and behaves oddly under impact. There is no material that wins on every count, because no such material exists for any real part. Material selection is the practice of turning "light, stiff, strong, cheap" into a single number for each candidate material, so that "best" stops being an opinion and becomes something you can calculate and defend.

The full treatment

First look: the bicycle frame problem

Lay five tubes of equal length and equal outer diameter on a bench, one each of steel, aluminum, titanium, wood, and carbon fiber composite. Load each as a cantilever and measure how far the free end deflects under the same weight. They deflect by very different amounts, because deflection depends on the material's stiffness, not just its shape. Now weigh each tube. The ranking by deflection and the ranking by mass are not the same ranking. A frame that never flexes but weighs the same as a small motorcycle is a solved problem nobody wants. The real question is never "which material is stiffest" or "which material is lightest," it is "which material gives the least mass for a given stiffness," and that is a different quantity entirely.

Building the idea: two clean properties, and one property built from both

Two measured properties do most of the work. Stiffness is how much a material resists elastic deformation under load, expressed as the Young's modulus E, which equals stress divided by strain (force per area divided by fractional deformation). Density is mass per unit volume. Neither number alone tells you what you need. What you need is specific stiffness, the ratio of stiffness to density, E divided by density, because that ratio answers the question "how much stiffness do I get per unit of mass I am forced to carry." A parallel ratio, specific strength, is strength (the stress at which the material yields or breaks) divided by density, and it answers the same question for load-carrying capacity instead of stiffness.

Building the idea: deriving why the ratio is the right number

This is not an arbitrary convenience; it falls out of the geometry of the part. Take the simplest case, a tie rod that must stretch by no more than a fixed amount under a fixed axial load. Stiffness of a rod under tension equals E times its cross-sectional area, divided by its length. To meet a target stiffness with a fixed length, the required area is proportional to load divided by E. The rod's mass equals density times area times length, so substituting the required area shows that mass is proportional to density divided by E, times a block of numbers fixed by the requirement (load and length) that no material choice can change. Minimizing mass for a fixed stiffness requirement therefore means minimizing density divided by E, which is the same as maximizing E divided by density, the specific stiffness. Run the same steps for a rod that must carry a fixed load without exceeding its strength rather than its stiffness, and the same logic produces specific strength, strength divided by density, as the quantity to maximize. This ratio is called a performance index: a single number, built from the geometry of the actual loading, that ranks materials correctly for that specific job. A different shape of part, a beam loaded in bending rather than a rod loaded in tension, produces a different index (the square root of E divided by density, for a beam of free thickness), because the mass-minimizing balance between stiffness and weight depends on how the material is arranged in space, not only on what it is made of.

Building the idea: charting and weighting multiple requirements at once

Real parts rarely have only one requirement. A bicycle frame must be stiff, strong, affordable, and tolerant of everyday impact, all at once. Two techniques handle this without pretending one number can capture everything. The first is charting: plot candidate materials on a log-log graph with stiffness on one axis and density on the other. A line of a given slope on that chart represents a constant value of the performance index, so materials lying on a higher such line genuinely outperform materials on a lower one for that specific loading case, regardless of how their raw stiffness or raw density compare in isolation. The second technique is staged screening and weighted scoring. First, screen out any material that fails a hard constraint, a maximum operating temperature, a minimum corrosion resistance, a manufacturing limit, no amount of stiffness or lightness compensates for a material that melts in service. Only then rank the survivors: normalize each remaining property onto a common scale (for instance, the value for each material divided by the best value among the candidates), assign a weight to each property reflecting how much the requirement actually matters for this part, and sum the weighted, normalized scores. The material with the highest sum is the best fit to the requirements as you have stated them, no more and no less.

Lineage

Choosing materials by feel is as old as toolmaking: a bronze age smith already knew that bronze held an edge that copper alone would not, and every craft tradition since has carried informal, hard-won knowledge about which material suits which job. What changed in the twentieth century was the sheer number of candidate materials. Once steel, several aluminum alloys, titanium, a growing family of engineering polymers, and fiber composites all competed for the same part, intuition built from apprenticeship stopped being enough to search the space reliably. Michael Ashby, working at Cambridge in the 1980s, systematized the practice into the property-chart and performance-index method taught here, building directly on the mechanics of materials that Gere, Timoshenko, and others had already formalized, and on the plain physical reasoning about strength and stiffness that J. E. Gordon set out for a general audience.

The strongest case for it

The method predicts real, checkable engineering shifts across very different industries. It correctly explains why aircraft structures moved from aluminum toward carbon-fiber composites as specific-stiffness requirements tightened, why bicycle frames moved from steel through aluminum toward titanium and carbon fiber for the same reason at a smaller scale, and why cheap, low-specific-strength materials remain entirely correct choices for parts where mass is not the binding constraint, a park bench has no reason to be built from titanium. Because the performance index is derived from the geometry of the loading rather than assumed, it generalizes: the same reasoning that ranks bicycle tubes ranks aircraft spars, furniture legs, and satellite struts, with only the index formula changing to match how the part is loaded.

The strongest case against it

The method's honesty depends on getting the index and the constraints right, and both are easy to get wrong. An index built for a rod in tension does not apply to a beam in bending or a shaft in torsion; using the wrong index silently reintroduces the very error, comparing materials as if only one property mattered, that the method exists to prevent. Weighted scoring across soft properties is only as good as the weights, and those weights are a judgment call, not a measurement, so two engineers can honestly disagree on the "best" material even after doing the analysis correctly. The method also says nothing by itself about manufacturability, joining, fatigue under repeated load, or long-term degradation, corrosion, creep, ultraviolet aging, unless those are explicitly added as screening constraints or extra properties; a chart that ranks only stiffness and density can recommend a material that later cracks at a weld or degrades in sunlight. A common misconception is treating the top-ranked material on a chart as automatically the correct choice, when in practice supply, cost volatility, and manufacturing constraints outside the chart routinely override a purely property-based ranking.

Where it stands now

The property-chart and performance-index method is standard, uncontested teaching in mechanical and materials engineering, and the underlying reasoning, mass minimization under a stated constraint, has not changed since its formalization. What has changed is tooling: searchable material property databases and selection software now let engineers screen thousands of candidate materials against dozens of constraints in place of the handful of tubes on a bench, but the logic each search performs is exactly the screening-then-ranking procedure described here.

Test yourself

You are asked to design a support strut for a solar panel on a small satellite. It must be stiff enough to survive launch vibration without resonating, light enough not to eat the spacecraft's mass budget, and able to survive repeated swings between minus 150 and plus 120 degrees without losing strength. First, decide whether the strut's dominant loading is closer to a tie rod in tension or a beam in bending, and write down the performance index that follows from that choice. Then list at least three candidate materials, screen out any that fail the temperature requirement outright, and rank the survivors using your index. Finally, state one real-world constraint, outside stiffness, density, and temperature, that could overturn your top-ranked choice, and explain why the chart alone could not have caught it.