The Assay

Seeds: fire assay and cupellation (14054), Charpy impact test (14055), proof testing of firearms (14056), bioassay (14057), non-destructive testing counter-case (14058), destructive testing thesis (14059). 6 source nodes across metallurgy, materials science, military history, pharmacology, and industrial inspection.

The word assay comes from Old French essai, from Latin exagium — a weighing. The word essay comes from the same root. Both mean trial. The difference is what gets consumed in the trying.

Cupellation is the oldest precision method for determining the purity of precious metals. Archaeological evidence at Fatmali Kalecik in Anatolia dates the technique to the fourth millennium BCE. Fuse the sample with lead in a bone-ash cupel at temperatures above 1,000 degrees Celsius. The lead oxidizes to litharge, absorbs the base metals, and is wicked into the porous cupel walls. What remains on the surface is a bead of precious metal. Weigh the bead. Compare it to the original sample. The difference is the impurity.

The accuracy is one part in ten thousand.

The non-destructive alternative existed alongside it for centuries. The Lydian touchstone — basanite, a fine-grained black siliceous stone — was in use by the sixth century BCE. Rub the gold against the stone. Compare the streak color to reference samples. Treat with nitric acid to heighten contrast. The touchstone is accurate to ten or twenty parts per thousand. One hundred times less precise than cupellation. But the gold survives the test.

The word hallmark means the mark applied at the Hall. In 1477, the Goldsmiths' Company of London established a permanent assay office at Goldsmiths' Hall, where all gold and silver was tested and stamped. The mark that certified the metal's purity was literally the mark of the building where the metal was destroyed to prove it was pure.

On the night of January 16, 1943, the T2 tanker Schenectady lay moored at the outfitting dock of Swan Island Shipyard in Portland, Oregon. She had completed her sea trials. The air temperature was minus six degrees Celsius. The water was calm. At approximately 11 p.m., the hull cracked almost in half just aft of the superstructure, the sound audible for at least a mile. The crack reached down both sides nearly to the keel. The keel fractured, and the bow and stern sagged to the river bottom.

The steel had passed all its tensile tests.

The United States built 2,710 Liberty ships during the war, using all-welded construction for speed. Of these, approximately 1,289 suffered brittle fracture damage. Twelve broke completely in two. The problem was not the welds. Constance Tipper, working at Cambridge, obtained pieces of the failed ships and demonstrated that the steel itself became brittle at low temperatures — it shattered like glass rather than bending. She showed that the ductile-to-brittle transition existed in the steel's own structure, invisible to any test that did not involve sudden impact at the relevant temperature.

Georges Charpy had published the test in 1901. A notched bar, 55 millimeters long by 10 by 10, with a V-notch 2 millimeters deep. A pendulum of known mass, raised to a known height, released to strike the bar in a single swing. The bar snaps or bends. The energy difference between the pendulum's height before and after the strike measures the energy absorbed. The specimen is destroyed. It cannot be tested again.

The Charpy test reveals what tensile testing cannot: the temperature at which ductile steel becomes brittle glass. The transition is abrupt. A few degrees separate the material that bends from the material that shatters. And the only way to find that boundary is to break the sample at every temperature until the curve appears. Today, every nuclear reactor pressure vessel contains surveillance capsules — small Charpy specimens, each the size of a finger, irradiated alongside the vessel for decades. When a capsule is withdrawn, each specimen is destroyed in a single swing. The entire operating license of a multi-billion-dollar plant depends on whether those bars snap or bend.

The Worshipful Company of Gunmakers received its Royal Charter from Charles I in 1637, though the first proof house was not built until 1657, near Aldgate. The process: load the firearm with proof ammunition generating pressures twenty-five to thirty percent above maximum service pressure. Fire it. If the barrel survives, stamp a proof mark into the metal. If it does not survive, the barrel has saved whoever would have fired it next.

The proof mark is not a guarantee that the weapon will function. It is a guarantee that the weapon has been subjected to conditions worse than it will encounter in service and has not failed. The test is the use case, run to excess. The etymology is the same: proof means trial. The proof of the pudding — William Camden recorded the proverb in 1605 — is in the eating. You cannot know whether a pudding is safe until someone eats it.

The same logic governs proof spirit. Before Bartholomew Sikes's hydrometer replaced it in 1816, the standard test for alcohol strength was to mix the spirits with gunpowder and ignite the mixture. If it burned with a steady blue flame, the spirit was at proof strength. If it would not ignite, under proof. The spirit was tested by partially consuming it in fire, and the fire was the answer.

Before chemical assay methods existed, the measuring instruments for drugs were alive.

In 1921, Frederick Banting and Charles Best isolated insulin at the University of Toronto. The extract was impure, variable between batches, and lethal in uncertain doses. The only way to standardize it was biological: inject the preparation into a two-kilogram rabbit. If the rabbit's blood glucose dropped by forty-five percent, the dose was one unit. The rabbit's convulsion was the measurement. The animal was the instrument.

Digitalis presented the same problem. William Withering published his observations on foxglove in 1785, but the active compound varied with the plant's stage of bloom, soil, drying method, and storage. In 1910, Hatcher and Brody developed the cat unit: infuse digitalis extract intravenously into a cat at a constant rate until cardiac arrest. The dose per kilogram of body weight that stops the cat's heart is the unit of measurement. Tablets were produced in strengths of one cat unit and two cat units. The method was used into the 1970s, until liquid chromatography replaced it.

John Scott Haldane investigated the Tylorstown Colliery explosion of January 1896 in Wales. He determined that miners were killed by carbon monoxide after the blast, not by the explosion itself. He recommended that miners carry a canary into the mines as an early warning sentinel. Canaries are approximately twenty times more sensitive to carbon monoxide than humans, showing distress roughly twenty minutes before a person would be affected. The canary's collapse was the measurement. But miners designed a resuscitation cage — a sealed glass-and-metal container with an oxygen valve on top. When the canary fell, the miner sealed the cage and opened the oxygen. The sentinel could be saved. The instrument that had just measured the danger could, sometimes, survive the measurement.

The last canary was retired from British mines in December 1986.

The counter-case is the entire field of non-destructive testing. Floyd Firestone patented the Supersonic Reflectoscope in 1942 — the first practical pulse-echo ultrasonic flaw detector, sending a 2.25-megahertz pulse through metal and reading the echoes. X-ray radiography for industrial inspection began in the 1920s. Eddy current testing, developed by Friedrich Forster at the Kaiser-Wilhelm Institute during the Second World War, uses electromagnetic induction to detect surface flaws. None of these methods destroy the specimen. All of them measure something different.

Non-destructive testing detects defects. It finds cracks, voids, inclusions — it answers the question is this piece probably acceptable? It does not answer the question what is this made of? It cannot measure fracture toughness, yield strength, ductile-to-brittle transition temperature, or exact alloy composition. For those, you still need to break something.

Every non-destructive method has a Probability of Detection curve — a function plotting the likelihood of detecting a flaw against flaw size. The standard metric is a90/95: the flaw size detectable with ninety percent probability at ninety-five percent confidence. Below that threshold, flaws are invisible. The POD curve is not a limitation of the instrument. It is a mathematical acknowledgment that seeing without touching has a resolution below which the method is blind. The only way to know what is below the threshold is to break the thing open.

Cupellation and the touchstone. The Charpy test and the tensile test. Proof firing and inspection. Bioassay and chromatography. Destructive testing and non-destructive testing. In each pair, the destructive method answers a harder question with greater certainty. The non-destructive method answers an easier question — is this probably acceptable? rather than what is this exactly? — and preserves the specimen. The two methods do not measure the same thing. The choice between them is the choice between knowing and keeping.

Every replacement of a destructive test by a non-destructive one succeeds by redefining the question. Sikes's hydrometer replaced the gunpowder test not by measuring proof strength more accurately, but by measuring specific gravity — a proxy for alcohol content that happens to correlate perfectly. Chromatography replaced the cat unit not by assaying the drug's biological effect more precisely, but by measuring its chemical composition — a different property that happens to predict the effect. The canary was replaced by an electronic sensor that detects carbon monoxide concentration — a physical measurement that correlates with biological danger without requiring the biology.

The progress is real. The animal no longer dies. The specimen no longer breaks. But the underlying structure persists: the more precisely you need to know what something is — not what it resembles, not what it correlates with, not whether it is probably acceptable — the more of it you must consume in the knowing. The touchstone approximates. The cupel tells the truth. And between the streak on the stone and the bead on the cupel lies a hundredfold gain in precision and the total destruction of the sample.