Industries / Mining and materials / Steel and metals / Precious metals refining
Assaying
Assaying is measuring how much gold, silver or platinum-group metal a piece of ore, scrap, coin or bar holds, so that miners, sellers and refiners can agree what it is worth.
5
Milestones achieved, from c. 550 BCE to 2022.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
c. 550 BCE
Streaks on a black stone
A touchstone is a fine, dark stone. An assayer rubs a piece of gold across it and compares the streak with streaks made by needles of known gold and silver content. One was found in a Lydian house at Sardis, dated to about the middle of the sixth century BCE. Its excavators write that a practiced jeweler can read a streak to better than 2 to 3%, and that touchstones are still widely used to test gold. Pliny the Elder, quoting Theophrastus, called it the "Lydian" stone.
Credit
Lydian goldworkers; the Archaeological Exploration of Sardis (Harvard), object S85.009
1282
Coins put on trial
The Trial of the Pyx checks that coins from the Royal Mint hold the gold and silver the law requires. The Bank of England's museum dates it to 1282; the Goldsmiths' Company dates the first public trial to 1248, when twelve London citizens and twelve goldsmiths examined the money. A jury still meets every year, at Goldsmiths' Hall since the Coinage Act of 1870. The Mint brings one coin from every batch it struck, normally more than 50,000 coins, and the Company's Assay Office tests a selection against reference metal plates called trial plates.
LessonThe test itself can be wrong; the 1710 trial judged the gold coins of Isaac Newton, then Master of the Mint, below standard, a verdict the Bank of England says was later proven mistaken.
Credit
Edward I; the Goldsmiths' Company and its Assay Office; the Royal Mint
1556
Assaying written down
Georgius Agricola's De Re Metallica, first printed in Latin in 1556, gave a whole book, Book VII, to assaying. An assay, he wrote, differs from smelting only in the small amount of ore used: it is melted in a little furnace exactly as the large furnace would do it, ending with gold and silver parted from lead in a cupel, a small dish made of ashes that soaks up the lead. Herbert and Lou Henry Hoover, who translated the book in 1912, write that the first explicit records of assaying are anonymous German booklets of the early 16th century, the Probierbüchlein, and that Agricola's account was a very great advance on them. The same small melt, now called fire assay, is still what refineries use to credit their customers for gold, as the refiner AG Metals puts it.
Credit
Georgius Agricola; translated by Herbert and Lou Henry Hoover
2012
Bars with tungsten hearts
In September 2012 Ibrahim Fadl, who runs a business near 47th Street in Manhattan, drilled into four 10-ounce gold bars he had bought for $100,000 and found grey tungsten inside, Fox News reported. Tungsten is nearly as dense as gold, so a real bar that has been hollowed out and filled with tungsten keeps its serial number and papers and weighs almost the same, NBC News reported, citing the New York station WNYW.
LessonWeighing a bar and testing its surface cannot see inside it; LBMA says there is still no non-destructive test it can endorse, and that only destructive testing of samples gives a bar's true fineness, its share of pure gold.
Credit
Ibrahim Fadl, who found them; reported by Fox News and NBC News
2022
A mine drops fire assay
PhotonAssay measures the gold in a jar of crushed rock by hitting it with high-energy X-rays that briefly activate the gold atoms, so nothing is melted and no lead or acid is used. It was developed at CSIRO, Australia's national science agency, from the early 2000s, with James Tickner as lead inventor, and is sold by Chrysos Corporation, formed in 2016. A jar holds about 400 to 500 g of rock, against 25 to 50 g for a fire assay, which matters where gold comes in coarse grains, as Tremblay, Tickner and colleagues reported in 2019. At Agnico Eagle's Fosterville gold mine in Victoria, Australia, 45,000 samples assayed both ways in 2020 and 2021 agreed closely but showed fire assay reading high above 250 grams per metric ton, and the mine switched to PhotonAssay in June 2022, its geologist Damon Symons said in 2024.
LessonAdoption took about five years of assaying every trial sample twice, at double the cost, and the mine's own risk assessment first found flaws in how it checked the new method.
Credit
James Tickner and CSIRO; Chrysos; the Fosterville mine
Watch for
Where the jar counts most
Robert Oliver and Dominique François-Bongarçon placed single weighed gold fragments at set points inside filled PhotonAssay jars and assayed each many times. The center of the jar gave the highest response and the sides were under-weighted, so they conclude that coarse-gold assays may be biased, because heavy gold grains can settle unevenly while a jar is filled. In the same issue James Tickner of Chrysos and colleagues reviewed field results and concluded that the effect adds only modestly to total uncertainty.
Achieved whena second laboratory reproduces the response map on another PhotonAssay unit, or paired assays from a coarse-gold mine show the bias the map predicts.
Claimed by
Oliver and François-Bongarçon, in Sampling Science & Technology, June 2026
Light and sound for tungsten
N. I. S. Abd Ghani, J. Jamaludin and colleagues propose photoacoustic sensing to find tungsten hidden in gold bars: laser light sets off sound waves in the bar, and the response differs where tungsten sits under the gold. The method has been modeled in COMSOL Multiphysics, a simulation program, not built.
Achieved whena working instrument finds tungsten inserts in real bars in a blind test run by someone else.
Claimed by
the authors, in the Malaysian Journal of Science Health & Technology, 2026
Open
01
Handheld XRF, tested blind
A public, blind comparison of handheld X-ray fluorescence (XRF) analyzers, which read the metals at an item's surface, against fire assay on at least 200 pieces of scrap gold jewelry. Include plated and gold-filled pieces, which have a gold layer over base metal. Each piece is read by three or more analyzer models before an accredited laboratory fire-assays it.
The briefCredit · unclaimed02
Bar checkers, tested blind
A blind test of the non-destructive checks dealers use on gold bars (electrical conductivity meters, ultrasonic thickness gauges, handheld XRF, and scales with calipers), run on genuine bars and on unmarked look-alike blocks with tungsten cores of different sizes. Publish, for each device and core size, how often it caught the core and how often it raised a false alarm on a genuine bar.
The briefCredit · unclaimed
Open 01 · the brief
Handheld XRF, tested blind
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, blind comparison of handheld X-ray fluorescence (XRF) analyzers, which read the metals at an item's surface, against fire assay on at least 200 pieces of scrap gold jewelry. Include plated and gold-filled pieces, which have a gold layer over base metal. Each piece is read by three or more analyzer models before an accredited laboratory fire-assays it.
Why it is still open
The published comparisons are small and mostly by makers. Thermo Fisher tested six large flat items and some small irregular ones, reported agreement with fire assay within 0.1 to 0.2% and better than 0.5% by weight, and advises against relying on portable XRF for gold-filled jewelry, bars and bullion. AG Metals, a refiner, compared 15 samples on one Niton model and recommends XRF only as a first check. Searches of OpenAlex and the web in November 2026 found only small single-laboratory or maker studies, none blind and none across several instrument models.
Built with what exists
Rented handheld XRF analyzers such as the Thermo Scientific Niton XL5 Plus, Evident Vanta and Bruker S1 TITAN; pieces lent by a scrap-gold buyer; a laboratory accredited to ISO/IEC 17025 that runs fire assay; Python with pandas for the statistics. The team melts nothing; the laboratory does the fire assay.
First step
Ask one scrap-gold buyer to log readings from its own analyzer on 50 incoming pieces, and price fire assays for the same 50 at an accredited laboratory.
Done looks like
At least 200 pieces read on at least three analyzer models, with each model's error in parts per thousand of gold published by karat band and by type of piece, and the share of plated or gold-filled pieces each model failed to flag.
Open 02 · the brief
Bar checkers, tested blind
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A blind test of the non-destructive checks dealers use on gold bars (electrical conductivity meters, ultrasonic thickness gauges, handheld XRF, and scales with calipers), run on genuine bars and on unmarked look-alike blocks with tungsten cores of different sizes. Publish, for each device and core size, how often it caught the core and how often it raised a false alarm on a genuine bar.
Why it is still open
LBMA says "There is currently no definitive non-destructive testing solution available that we can endorse", and that the only accurate way to find a bar's fineness is destructive testing of representative samples. New methods are still at the simulation stage, such as the 2026 photoacoustic model under Watch for, and makers of ultrasonic gauges describe the use in application notes without detection rates. Searches of OpenAlex and the web found reviews, vendor claims and single-method studies, but no blind tally of the devices dealers use.
Built with what exists
Devices sold to dealers, such as the Sigma Metalytics Precious Metals Verifier (conductivity) and an ultrasonic thickness gauge such as the Evident 38DL PLUS; a handheld XRF analyzer; a protocol registered in advance on the Open Science Framework (osf.io). The genuine bars and the look-alikes are made, held and tested by a licensed assay office or refiner; the look-alikes carry no brand, serial number or hallmark, never leave the laboratory and are melted down afterwards.
First step
Write the protocol and ask two assay offices, or LBMA's referee laboratories, whether one will host the test and supply the bars.
Done looks like
At least 40 pieces (20 genuine bars and 20 look-alikes with cores from 10% to 60% of their volume), five or more device types, each piece tested blind by two operators, and detection and false-alarm rates published for every device and core size.